Voltage-Gated Ion Channels Mediate the Electrotaxis of Glioblastoma Cells in a Hybrid PMMA/PDMS Microdevice Hsieh-Fu Tsai,1 Camilo Ijspeert,1 and Amy Q

bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Voltage-gated ion channels mediate the electrotaxis of glioblastoma cells in a hybrid PMMA/PDMS microdevice Hsieh-Fu Tsai,1 Camilo IJspeert,1 and Amy Q. Shen1, a) Micro/Bio/Nanofluidics Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa JAPAN (Dated: 18 February 2020) Transformed astrocytes in the most aggressive form cause glioblastoma, the most common cancer in central nervous system with high mortality. The physiological electric field by neuronal local field potentials and tissue polarity may guide the infiltration of glioblastoma cells through the electrotaxis process. However, microenvironments with multiplex gradients are difficult to create. In this work, we have developed a hybrid microfluidic platform to study glioblastoma electrotaxis in controlled microenvironments with high throughput quantitative analysis by a machine learning-powered single cell tracking software. By equalizing the hydrostatic pressure difference between inlets and outlets of the microchannel, uniform single cells can be seeded reliably inside the microdevice. The electrotaxis of two glioblastoma models, T98G and U-251MG, require optimal laminin-containing extracellular matrix and exhibits opposite directional and electro-alignment tendencies. Calcium signaling is a key contributor in glioblastoma pathophysiology but its role in glioblastoma electrotaxis is still an open question. Anodal T98G electrotaxis and cathodal U-251MG electrotaxis require the presence of extracellular calcium cations. U-251MG electrotaxis is dependent on the P/Q-type voltage-gated calcium channel (VGCC) and T98G is dependent on the R-type VGCC. U-251MG and T98G electrotaxis are also mediated by A-type (rapidly inactivating) voltage-gated potassium channels and acid- sensing sodium channels. The involvement of multiple ion channels suggests that the glioblastoma electrotaxis is complex and patient-specific ion channel expression can be critical to develop personalized therapeutics to fight against cancer metastasis. The hybrid microfluidic design and machine learning-powered single cell analysis provide a simple and flexible platform for quantitative investigation of complicated biological systems.

I. INTRODUCTION electric fields in the brain may play an important role in mediating the glioma tumorigenesis and invasion15–19. Glioma is one of the most common types of brain Cells sense the electric field by bioelectrical activation of cancer and the aggressive form of it, glioblastoma, con- voltage-sensitive proteins, mechanosensing due to elec- tributes to poor prognosis, high mortality, and high prob- trokinetic phenomena, or activated chemical signaling ability of recurrence1,2, due to the infiltration nature of due to electrokinetically polarized membrane receptors the disease. The highly infiltrative ability of glioblas- (Supplementary FIG. S.1). The voltage gradient cre- toma originates from the invasive/migratory ability of ates a large voltage drop at cellular membrane which glioma stem cells or brain tumor initiating cells3,4. Not can directly activate voltage sensitive proteins such as only glioma cells are important, but also the microen- voltage-gated ion channels that are most commonly ex- vironment in the brain helps shaping the heterogeneity pressed on excitable membranes at neuronal synapses 20 of the glioma5. The glioma cells interact with the ex- and neuro-muscular junctions . Among the voltage- tracellular matrix (ECM), glial cells, and immune cells gated ion channels in the brain, calcium channels are in the brain and mediate the formation of peri-vascular, especially important as calcium influx plays a pivotal 21,22 peri-necrotic, and invasive tumor microenvironments6–9. role in cellular signaling . Calcium signaling is also Understanding the molecular mechanisms of the invasive- important in glioma cell proliferation, resistance to ther- 23–27 ness in glioma cells with respect to the tumor microenvi- apy, and metastasis . Whether or not the calcium ronment is vital for developing new therapeutic options signaling in glioma is mediated by electric field is still an and improving patient outcome10,11. open question. In the brain, glial cells are immersed in an electric field Conventional in vitro electrical stimulation systems created by tissue polarity from brain macrostructures as for studying cell responses in electrical microenvi- well as the local field potentials which are established ronments are bulky and the experimental through- from the action potentials fired by the neurons12. A weak put is limited28,29. To overcome these limitations, endogenous electric current has been shown to serve as a robust high-throughput in vitro platform that cre- a guidance cue for neuroblast migration from the sub- ates stable electrical stimulation of cells and inter- ventricular zone in mouse13, a region speculated as the faces with automated microscopy is a prerequisite for origin of glioma tumorigenesis14. Thus, the physiological rapid screening of targets and identifying molecular mechanisms. To this end, we have developed a hybrid poly(methylmethacrylate)/poly(dimethylsiloxane) (PMMA/PDMS) microfluidic platform to reliably study a)[email protected] glioblastoma single cell migration under high-throughput bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 2 dcEF stimulations that multiple antagonists can be tested simultaneously to identify molecular mechanisms. Quantitative single cell migration analysis is carried out by extracting cell migration metrics such as the directedness, orientation, or speed using a robust machine learning-powered cell segmen- tation/tracking/analysis software with stain-free phase contrast microscopy30. Using the hybrid microfluidic platform, the role of voltage-gated calcium channels in calcium signaling pathways of glioblastoma electrotaxis are investigated.

II. RESULTS

A. Uniform single cell seeding by submerged manipulation in hybrid multiple electric ﬁeld chip (HMEFC)

To analyze single cell migration in microchannels, cells must be seeded sparsely and allowed to adhere and cul- FIG. 1. The results of U-251MG cell seeding inside mi- ture reliably31,32. However, it is known that uneven dis- crochannels by various methods. (a, d, g, j) In tip loading tribution of cells due to fluid flow, convection in suspen- method, cells are introduced by using gravitational flow with sion and vessel movement after seeding can cause aggre- micropipette tips. The cells can flocculate inside the tips and gation and differentiation of cells33. Different cell loading in microchannels as illustrated in (d). The microscopy image of seeded cells is shown in (g) and magnified in (j); (b, e, h, methods affecting cell distribution in single cell migration k) In tip injection method, cells are injected into the channels experiments are investigated, such as tip loading method, and tips are removed. The small hydrostatic pressure differ- tip injection method, and a pressure-balanced submerged ences between the inlet/outlet (shown as ∆h) will contribute cell seeding, as shown in FIG. 1. to hydrodynamic flow and disturb the cell distribution, caus- In tip loading method (FIG. 1.a), the cells flocculate ing non-uniform cell distribution and aggregates as shown in (e). The microscopy image of seeded cells is shown in (h) in the small pipet tip and cannot be dispersed uniformly and magnified in (k); (c, f, i, l) In our pressure-balanced sub- in the microchannel (FIG. 1.d, g, j). In tip injection merged cell seeding method, the hydrostatic pressure differ- method (FIG. 1.b), the cells are originally injected in the ence is eliminated. The injected cells remain uniform through- channels with uniform cell distribution. However, without the channel as shown in (f). The microscopy image of out balancing the microchannel inlet/outlet pressure, the seeded cells is shown in (i) and magnified in (l). The uniform minute hydrostatic pressure difference between inlets and and sparse cell seeding method is suitable for many different outlets generates a small pressure-driven flow that dis- applications from single cell tracking, ensembled cell studies places cells which lead to cell aggregates (FIG. 1.e, h, k). to cell assembly. The scale bars in (g, h, i) represent 500 µm. Furthermore, in tip injection method, due to the small The scale bars in (j, k, l) represent 200 µm. dimension of the punched holes at inlet/outlet interfaces, bubbles are easily trapped and may be introduced into microchannels, disrupting fluid advection and chemical B. Glioblastoma electrotaxis requires optimal transport. laminin-containing ECM By submerging inlets and outlets underwater using a reversibly bonded top reservoir and balancing the pres- An effective ECM coating on the substrate is essen- sure between inlets and outlets (FIG. 1.c), air bubbles can tial for cell adhesion and formation of focal adhesions be avoided and pressure-driven flow is prevented from af- for cell migration34,35. Glioblastoma can be molecu- fecting cell distribution. Moreover, using this cell seeding larly classified into proneural, neural, classical, and mes- method, only minute amount of cells is needed (the vol- enchymal types according to The Cancer Genome Atlas ume of the microchannel). Uniformly distributed single (TCGA)36. We use two glioblastoma cell models, T98G cell seeding across the entire microchannel is obtained and U-251MG, which are both of caucasian male ori- for single cell migration experiments (FIG. 1.f, i, l). The gin and classified as mesenchymal type with p53 mutant top reservoir in our submerged cell seeding setup can be genotype37,38. The adhesion and electrotaxis of T98G easily removed after cells are seeded and can be adapted and U-251MG glioblastoma cell lines on various ECMs to a wide range of microfluidic chips. are tested in a double-layer hybrid multiple electric field bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 3 chip (HMEFC) based on the hybrid PMMA/PDMS de- C. Electrotaxis behavior may reflect the heterogeneity of sign approach17 (Supplementary TABLE. S.1). The mi- glioblastoma gration directedness, speed, and morphology of glioblastoma cells (see details in section IV.D) are quantitatively To further analyze how T98G and U-251MG cells mi- analyzed by a machine learning-based single cell segmen- grate under dcEF stimulation, the glioblastoma electro- tation and tracking software from stain-free phase con- taxis in serum-containing (FBS) and serum-free media 30 trast microscopy . The cell morphologies of the two were examined. cell lines on various ECMs are shown in Supplemen- a. T98G and U-251MG cells migrate toward opposite tary FIG. S.2. directions under dcEF stimulation FIG. 2 shows the di- While standard poly(D-lysine) (PDL) and various rectedness and speed of T98G and U-251MG electrotaxis. combinations of poly(L-ornithine) (PLO) and laminin Upon 300 V m−1 dcEF stimulation, T98G cell migrate have been used for glioblastoma electrotaxis16,39, the ad- toward anode (positive electrode) while the U-251MG hesion and electrotaxis of T98G and U-251MG are not al- migrate toward cathode (negative electrode). Further- ways are not always consistent and reproducible as shown more, the directedness in the electrotaxis of T98G cells in Supplementary FIG. S.2 and FIG. S.3. T98G and does not depend on the presence of fetal bovine serum U-251MG electrotactic responses are also not stably re- (FBS) (P = .45) but the speed is significantly decreased produced on collagen I, collagen IV, vitronectin, and fi- (P < .0001). The directedness of U-251MG cell electro- bronectin coatings. taxis however is highly dependent on the presence of FBS Both T98G and U-251MG cells adhere well and demon- (P < .0001) and lack of FBS also decreases the speed of strate lamellipodia structures on substrates containing U-251MG electrotaxis (P < .0001). The FBS serum is laminins, such as pure laminin coating, or GeltrexTM. rich in growth factors, proteins, and ions, which can en- GeltrexTM is a growth factor-reduced complex basement hance the chemical signaling in electrotaxis (Supplemen- membrane extract purified from murine Engelbreth- tary FIG. S.1). The electrotaxis and random migration Holm-Swarm tumors containing laminin, collagen IV, en- of T98G and U-251MG with or without electrical stimu- tactin, and heparin sulfate proteoglycan40. Cells interact lation is shown in Supplementary VIDEO. S.1 – Supple- with laminins through various integrins including α1β1, mentary VIDEO. S.4. α2β1, α3β1, α6β1, α7β141,42. The integrins are believed Both T98G cells and U-251MG cells are categorized 37,38 to participate in the initiation of electrotaxis through as mesenchymal type glioblastoma , however, their mechanosensitive pathways43–45. electrotactic responses are completely different. Similar results are reported in the electrotaxis of glioblastoma Discussed in details in section C, under electrical stim- cells and spheroids15,16,19,39 and lung adenocarcinoma51, ulation, the electrotaxis of both T98G and U-251MG are showing that although molecular and surface marker more prominent and reproducible on GeltrexTM coatings, makeups of the cell lines are similar, their electrotaxis hence all the studies in the following sections are based responses can be completely different. The opposite on GeltrexTM coatings. The detailed data of T98G and electrotaxis results may reflect the fundamental hetero- U-251MG electrotaxis on various ECMs is shown in Sup- geneity among glioblastoma cells which has been spec- plementary TABLE S.2. ulated to contribute to the recurrence and therapeutic An interesting observation is found in U-251MG resistance after anti-tumor therapy52–56. Further eluci- cells on iMatrix-511-coated substrates. U-251MG cells dation of the correlation among electrotactic responses, demonstrate large lamellipodia associated with high metastatic properties of glioblastoma, and in situ elec- migratory speed (15.45 µm hr−1 under 300 V m−1, tric field around the lesion may be beneficial to evaluate P < .0001) but with diminished directedness (0.01, electrotaxis response as a predictive tool for glioblastoma P < .0001). Note that iMatrix-511 is a recombinant trun- metastasis. cated laminin with α5β1γ1 subunits and interacts with b. Only T98G cells demonstrate prominent electro- 46,47 cells through the α6β1 integrin . This suggests that alignment behavior under electrical stimulation Aside the specific molecular configuration of laminins in ECM from directional migration in the dcEF, cells may also may be vital for electrotaxis. demonstrate long axial alignment in perpendicular to While it is understood that ECMs in tumor microen- the dcEF vector. While this phenomenon is commonly vironment are important48, glioblastoma cells demon- observed among many cell types51,57–60, the molecular strate preference for adherence and electrotaxis on mechanism and the biological role are not clear. Electro- laminin-coated surfaces. Within the brain microenvi- alignment may participate in the cytoskeletal restructur- ronment, laminin expressions are restricted to the base- ing in tissue morphogenesis61,62, but biophysical studies ment membrane of neural vasculature49 and perivascular show that in vitro microtubules align in parallel to elec- tumor microenvironment is especially vital for glioblas- tric field vectors rather than perpendicular63,64. toma metastasis10,50. Therefore, the correlation among Supplementary FIG. S.4 shows orientation of the cells laminin, glioblastoma electrotaxis, and the perivascular with respect to time in electrically stimulated T98G and invasion process may be important in glioblastoma can- U-251MG cells over 6 hours. The orientation index is cer biology that requires further elucidation. defined as the average cosine of two times the angle be- bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 4

FIG. 2. The electrotaxis of T98G & U-251MG glioblastoma cells in dependence of the extracellular serum and calcium by varying the medium recipe. (a) The electrotaxis directedness; (b) The electrotaxis speed; n.s. indicates not significant; **** indicate P < .0001; The numbers in parentheses indicate actual P-values. FBS: DMEM with serum; SF: serum-free DMEM ; CaFree-FBS: 0 mM calcium DMEM with 10% FBS; CaFree-SF: 0 mM calcium serum-free DMEM. tween the long axis of a cell and the electric field vector cium ion in cell culture media may impact calcium home- (See more details in Supplementary FIG. S.11 & Supple- ostasis significantly but the loss of viability is rescued mentary TABLE. S.5). For a group of perpendicularly by the addition of 10% FBS which contains calcium and oriented cells, the average orientation index is -1 and for growth factors. The electrotaxis of U-251MG in calcium- a group of parallely orientated cells, the average orienta- free, serum-free media is not affected compared to those tion index is 1. For a group of randomly arranged cells, in serum-free media (P > .99). Interestingly, the elec- the average orientation index should be 0. Only T98G trotactic speed of U-251MG in calcium-free serum-free cells show prominent perpendicular alignment after elec- media further increases (P < .0001). trical stimulation (Indexorientation at 0 hr v.s. 6 hr is -0.11 Second, to validate that calcium cations are important, v.s. -0.83, P < .0001). FBS deprivation slightly decreases cation chelators EDTA and EGTA are used to chelate the alignment tendency and delays the onset but does not the free calcium in the cell culture media with 10% FBS. abolish it (Indexorientation of serum free v.s. 10% FBS at Treatment of 2 mM EDTA significantly represses the di- 6 hr is -0.48 v.s. -0.83, P < .0001). However, U-251MG rectedness of only U-251MG cells (P < .0001, Supple- does not show any perpendicular alignment. These re- mentary FIG. S.6.a). To further confirm the electrotaxis sults further illustrate the heterogeneity of glioblastoma inhibition is due to extracellular calcium cations, EGTA, cells. The cell electroalignment phenotypes of T98G and a divalent cation chelator with increased affinity towards U-251MG after dcEF stimulation are shown in Supple- Ca2+, is used. At 1 mM EGTA, the electrotactic direct- mentary FIG. S.5. ednesses of neither cell lines are affected but the electrotactic speeds of them become more dispersed (P < .0001). Under 2 mM EGTA treatments, the electrotactic direct- D. Glioblastoma electrotaxis requires extracellular calcium edness of U-251MG cells are reduced (P < .0001). When 5 mM EGTA is used, the electrotaxis of both cell lines are further repressed in both directedness and speed and Calcium ion flux is known to be involved in the elec- the cells detach from the substrate. These results sug- trotaxis signaling of multiple cell types including mouse gest that glioblastoma electrotaxis requires extracellular fibroblasts, human prostate cancer cell, and neural crest calcium cations and calcium influx may be important for cells65–69. Deregulation of calcium influx in the cells re- electrotaxis, particularly that of U-251MG cells. duces actin polymerization and affects cell motility speed but its effect on electrotactic directedness vary depend- ing on cell types65,69. The hypothesis that glioblastoma electrotaxis is dependent of extracellular calcium cations E. Glioblastoma electrotaxis is mediated by voltage-gated is tested in this work (FIG. 2). ion channels First, a calcium-free, serum-free cell culture media (CaFree SF) is used to test if glioblastoma electrotaxis Ions channels expressed on glioblastoma cells includ- requires extracellular calcium. T98G cells lose viability ing various potassium, calcium, sodium, and chloride both with and without dcEF stimulation. Lack of cal- ion channels are believed to facilitate pathogenesis of bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 5

FIG. 3. The electrotactic directedness of T98G & U-251MG glioblastoma cells under 300 V m−1 dcEF after 6 hours with pharmacological inhibition on various ion channels. † indicates the electrotaxis tested against those without EF stimulation; ‡ indicate the electrotaxis group tested against those with cytochrome C which prevents adsorption of short peptides to experimental apparatus; All other groups are statistically compared to their respective controls in cell culture media with 10% FBS; n.s. indicates not significant; * indicates P < .05; ** indicate P < .01; *** indicate P < .001; **** indicate P < .0001; The numbers in parentheses indicate actual P-values. glioblastoma27,70–72. Although the expressions of numer- trotaxis of other cell types has been proposed. Trollinger ous ion channels vary among clinical glioma samples25, et al., find that a stretch-sensitive VGCC mediates the ion channel expression profiles have been suggested to electrotaxis of human keratinocytes80. Aonuma et al., predict survival in glioma patients73,74. Glioblastoma report that T-type VGCC mediates the electrotaxis of cells are immersed in the local field potentials within the green paramecia81. L-type VGCCs also regulate the brain, and extracellular calcium is required for glioblas- chondrogenesis during early limb development which is toma electrotaxis which can flow into the cells through known to be a bioelectricity process82. voltage-gated calcium channels (VGCCs). Whether and how VGCCs participate in the electrotaxis of glioblas- Another class of membrane proteins that are bio- toma may shed new insights for inhibiting glioblastoma electrically activated are voltage-gated potassium chan- infiltration. nels (VGKCs), which are represented by 12 families (Kv1–Kv12). VGKCs are involved in diverse physio- VGCCs are known to play important roles in glioma logical and pathological processess regulating the re- biology such as cell proliferation, apoptosis, and sen- polarization of neuromuscular action potential, calcium sitization to ionizing radiation27,75,76. VGCCs can be homeostasis, cellular proliferation, migration, and can- categorized as high voltage activated (HVA) or low cer proliferation83–89. Voltage-gated potassium channel 77–79 90 voltage activated (LVA) types . Among the HVA Kv1.2 and non voltage-gated inwardly-rectifying potas- 91 VGCCs, four subtypes can be categorized by electro- sium channel Kir4.2 have been shown to be involved in physiological property and genetic phylogeny, includ- the sensing of electric field and signaling of cell electro- ing L-type (long-lasting, Cav1.1–1.4), P/Q-type (purk- taxis. The potassium ion transporters confer biophysical inje/unknown, Cav2.1), N-type (neural, Cav2.2), and R- signals that are key for regulating stem cells and tumor 92 type (residual, Cav2.3) VGCCs. The LVA VGCCs are cells behavior in microenvironment . In prostate can- composed of T-type channels (transient, Cav3.1–3.3). Al- cer cells, VGKC expressions are linked to the increased 93 though the involvement of VGCCs in the electrotaxis of metastatic potential . Furthermore, inhibition of Kv1.3 glioblastoma is not yet elucidated, VGCCs’ role in elec- VGKC has been shown to induce apoptosis of glioblas- bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 6 toma cells in vitro94. trotaxis. However, U-251MG electrotaxis is not affected The role of VGCCs in glioblastoma cell electrotaxis when tested using another potent T-type VGCC in- is investigated by using pharmacological inhibitors (Sup- hibitor, TTA-A2108,109. plementary TABLE. S.3). Many of these inhibitors are c. T98G and U-251MG electrotaxis are also mediated short peptides purified or recombinant engineered from by A-type VGKCs and acid-sensing sodium channels In venoms of poisonous species. When testing inhibition, T98G electrotaxis, cadmium and SNX-482 inhibit the R- cytochrome C was added to the culture media to pre- type HVA VGCC and decreases electrotaxis directedness. vent non-specific adsorption of peptide inhibitors to mi- However, cadmium and SNX-482 have been reported to crofluidic chips95. The detailed results of pharmaco- also block a rapid inactivating (A-type) transient out- logical VGCC inhibition on the electrotaxis of T98G ward VGKC (Kv4.3) and experimental results of SNX- and U-251MG cells are shown in FIG. 3, Supplemen- 482 should be interpreted carefully110–112. Using 5 µM tary FIG. S.7, Supplementary FIG. S.8, and Supplemen- AmmTx3, a member of the α-KTX15 family of scorpion 113–115 tary TABLE. S.4. toxins, to block A-type VGKCs (Kv4.2 & Kv4.3) , In both glioblastoma cell lines, inhibition of L-type T98G electrotaxis directedness (FBS v.s. AmmTx3 = - HVA VGCC with gadolinium96 or nicardipine97 exhibit 0.69 v.s. -0.39, P = .0025) and speed (FBS v.s. AmmTx3 no effect on electrotactic directedness (P > .99) nor on = 9.99 v.s. 4.42, P < .0001) are repressed but not com- speed (P > .15) (Supplementary FIG. S.7). Inhibition pletely abolished (Supplementary TABLE S.4). Further- of N-type HVA VGCCs with ω-Conotoxin GVIA98,99 more the inhibition on T98G directeness caused by SNX- also has no effect on the electrotaxis of either cell type 482 is stronger than AmmTx3 (SNX-482 v.s. AmmTx3 (P > .92). = -0.27 v.s. -0.39, P = .039, two-tailed t test). An- a. T98G electrotaxis is mediated by R-type HVA other broad spectrum transient VGKC inhibitor, 4- VGCC The electrotaxis of T98G is repressed when aminopyridine (4-AP), was used to confirm the results treated with cadmium which is a broad spectrum HVA from AmmTx3116–118. Under 1 mM 4-AP, the directed- VGCC inhibitor at 50 µM and 100 µM (P < .0001, ness and speed in T98G electrotaxis were both repressed FIG. 3)100,101. Upon further identification, the directed- (P < .0001, FIG. 3). Increasing of 4-AP to 4 mM, though ness in T98G electrotaxis is repressed by use of SNX-482, results suggest that electrotactic directedness has not an R-type VGCC inhibitor102,103 (P = .049). The elec- changed (P > .99) compared to control, but the migra- trotaxis of T98G cells repressed with SNX-482 is shown tion speed is decreased (P < .0001). This is likely an arti- in Supplementary VIDEO. S.5. fact from part of T98G cells starting to detach from the Calmodulin, a calcium binding protein, mediates surface rather than actual electrotaxis (Supplementary many of the Ca2+ dependent-signaling by interact- FIG. S.9). These results suggest that T98G electrotaxis ing with VGCCs and maintaining intracellular calcium may be mediated by A-type VGKCs (Kv4.3), but the in- homeostasis104,105. However, the electrotaxis of T98G is volvement of R-type VGCCs cannot be completely ruled not dependent on calmodulin by inhibition with calmi- out. Further molecular studies into the roles of VGCCs dazolium (P > .99, FIG. 3) and Ni2+ treatment has no and VGKCs in glioblastoma electrotaxis are required. inhibition on T98G cells (P > .99) which has partial in- Similar inhibition of A-type VGKC also represses U- hibition on R-type VGCC103,106. These results imply an 251MG electrotaxis. When U-251MG cells were treated alternative mechanism might be at play. with AmmTx3 and 4-AP, the electrotaxis directedness is b. U-251MG electrotaxis is mediated by P/Q-type inhibited by both compounds (P < .0001) and the speed HVA VGCCs U-251MG electrotaxis has exhibited its is repressed in only 4-AP (P <.0001) but not AmmTx3 dependency on HVA VGCCs. Decreased directedness (P = .09) (Supplementary FIG. S.8). The electrotaxis of and speed are observed in U-251MG cells treated with T98G and U-251MG is suppressed when A-type VGKC 100 µM cadmium (P = .0372, Supplementary TABLE is inhibited through 1 mM 4-AP are shown in Supple- S.4). Upon further identification, U-251MG electrotaxis mentary VIDEO. S.7 & VIDEO. S.8). directedness is repressed when treated with P/Q-type Furthermore, nickel and amiloride repress the direct- HVA VGCC inhibitor using agatoxin IVA and conotoxin edness of U-251MG electrotaxis but not through T-type MVIIC (P < .0001) (FIG. 3). However, the electro- VGCC. At high concentration of nickel and amiloride, the tactic speed is not affected by agatoxin IVA (P > .99) compounds may inhibit acid-sensing ion channel (ASIC) but decreased by conotoxin MVIIC (P < .0001) (Sup- and epithelial sodium channel (ENaC), which are mem- plementary FIG. S.8). The electrotactic directedness of bers of a superfamily of voltage-insensitive mechanosensi- U-251MG is dependent on calmodulin (P < .0001). The tive sodium channels119,120. Furthermore, ASIC sodium electrotaxis of U-251MG is repressed by the treatment of channels are specifically expressed in the high-grade agatoxin IVA, shown in Supplementary VIDEO. S.6. glioma cells but not in normal brain tissues or low grade Furthermore, the electrotactic directedness of U- glial cells121,122. Sodium ion flux is known to medi- 251MG cells is repressed with nickel (P < .0001) and ate electrotaxis in keratinocytes through ENaC sodium amiloride (P < .01) as well as the electrotactic speed channels123 and prostate cancer cells through voltage- (P < .0001) (FIG. 3). These results suggest a possi- gated sodium channels124–126. ble involvement of T-type VGCC107 in U-251MG elec- To confirm the involvement of ASIC sodium chan- bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 7

diate glioblastoma electrotaxis is necessary to map the signaling network that may contribute to glioblastoma metastasis. The high-throughput hybrid microdevices and machine learning-assisted quantitative analysis developed in this work can be very useful for systematic phenotype proﬁling and identiﬁcation of molecular mechanisms underlying cell electrotaxis.

III. CONCLUSION

The hybrid PMMA/PDMS microfluidic chip is a ro- FIG. 4. The signaling mechanism identified in this study. bust platform for high throughput electrotaxis studies. Laminin-based ECMs is necessary for glioblastoma electro- Cell migration in multiple dcEFs under multiplex con- taxis, suggesting that integrins may play a role. The voltage- gated ion channels and ion transporters also mediate glioblas- ditions can be studied in a single experiment in combi- toma electrotaxis that requires extracellular calcium. nation with automated microscopy. The submerged op- eration balancing the inlet/outlet hydrostatic pressures guarantees a stable microenvironment that avoids mi- nels in U-251MG electrotaxis, U-251MG cells are treated crobubbles, ensures uniform cell seeding, and minimizes with 5 µM benzamil hydrochloride (Alomone labs, required number of cells. Uniformly distributed single USA)127–129. The directedness of U-251MG electrotaxis cells can be reliably seeded in microfluidic chip that fur- is significantly repressed by benzamil treatment (FBS ther increases the robustness for high throughput and v.s. benzamil = 0.58 v.s. 0.29, P < .0001, FIG. 3) but more reproducible experiments. Use of machine learning- not the speed (P = .42) (more details in Supplementary enabled single cell migration analysis automates the cell FIG. S.8). Benzamil is also tested on T98G electrotaxis migration data analysis workflow for reliable quantitative and found to inhibit its directedness (P < .01) and speed data at high throughput. (P < .0001). GeltrexTM coating has been identified to support re- d. Various ion channels participate in the electrotaxis producible electrotaxis model of T98G and U-251MG of glioblastoma cells of different origins The pharma- glioblastoma cells. The heterogeneity responses of T98G cological studies on the ion channels in T98G and U- and U-251MG electrotaxis and the importance of calcium 251MG electrotaxis suggest that multiple ion channels, signaling are identified. By further inhibitorial study, which may be voltage-sensitive or not, can mediate the the electrotaxis of T98G may depend on R-type HVA sensing of endogenous electric field and initiate the mi- 13 VGCCs, A-type VGKCs, and ASIC sodium channels. gratory response . The proposed mechanism is high- The electrotaxis of U-251MG depends on P/Q-type HVA lighted in FIG. 4. VGCCs, A-type VGKCS, and ASIC sodium channels. A-type VGKC, R-type VGCC, and ASIC sodium chan- Multiple ion channels, which may be voltage-sensitive or nels mediate the electrotaxis of T98G cells while P/Q- not, can mediate the sensing of electric field and electro- type VGCCs, A-type VGKC, and ASIC sodium channels taxis in different glioblastoma models, suggesting that mediate the electrotaxis of U-251MG cells. These results glioblastoma infiltration can be amplified by endogenous suggest that ion channel expression profiles are cell line electric field in a tumor sample-dependent manner. The specific and correlating ion channel expressions with elec- roles of ion channels on glioma metastasis and survival trotactic phenotypes of cancer cells may be beneficial to with regard to physiological electric field require further provide new insights of metastasis-aimed therapeutics by systematic studies and in vivo validation. inhibiting electrotaxis130,131. If glioblastoma infiltration can be inhibited by targeted therapeutics, the quality-of- As proof of principle, the hybrid PMMA/PDMS mi- life and prognosis of glioblastoma patients could be im- crofluidic design demonstrates robustness and versatility proved. The downstream molecular signaling of VGCCs, for high throughput experiments. The microfluidic chip VGKCs, and ASIC sodium channels in glioblastoma elec- design can be tailor-made for specific biological study. By trotaxis is an interesting future direction to investigate. using a robust, flexible, and high-throughput microflu- Recently, glutamatergic receptors have also been shown idic platform together with machine learning software, to mediate neuron-glioma interaction and glioma pro- the bottleneck of data analysis in high throughput ex- gression through calcium signaling132–134. Therefore, a periments can be resolved, opening new opportunities for systematic screening of potassium channel, sodium chan- quantitatively studying cell responses in well-controlled nels, and glutamate receptor ion channels’ ability to me- microenvironments. bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 8

IV. METHODS ity of 80. The numerically simulated electric field ratio was 4.99: 2.45:1:0 in section I to IV and section V to VIII A. Hybrid multiple electric field device (HMEFC) design, with limited chemical transport across the interconnect- simulation, and fabrication ing channels as designed (Supplementary FIG. S.13). To fabricate HMEFC, first, the PDMS chip was fabri- 137 To phenotypically test cell electrotaxis and elucidate cated by soft lithography technique . The double-layer molecular mechanism, a reliable in vitro platform for high microfluidic design was fabricated into a mold using neg- throughput study is necessary. The HMEFC is designed ative photoresists (10 µm and 100 µm) on a silicon wafer by the hybrid PMMA/PDMS approach17 (FIG. 5.a). Us- using direct-write lithographic writer and mask aligner ing the hybrid PMMA/PDMS approach, the prototyping (DL-1000, Nano Systems Solution, Japan and MA/BA6, disadvantages in both materials can be mitigated while SUSS MicroTec, Germany) . After passiviation of the the advantages can be combined. In PMMA, complex mold with perfluorosilane, mixed PDMS monomer (10:1 3D structures for fluidic routing or reservoir and world- monomer:curing agent ratio, Sylgard 184, Corning, USA) was poured and cured on the master mold in a custom- to-chip interface can be quickly prototyped by CO2 laser cutting and thermal bonding. However, spatial resolu- made casting block which ensures the 4 mm thickness tion using this approach is not sufficiently high to cre- in finished PDMS devices. After degassing, a piece of ate reliable microfluidic environments. In contrast, pre- 15 mm-thick PMMA block was placed on top of the cast- cise quasi-two dimensional microstructures can be fab- ing block to ensure the surface flatness of PDMS. The ricated using the soft lithography technique in PDMS PDMS was cured in oven and cut to yield individual de- microfluidic chips. But standard soft lithogarphy for vices. 1 mm-wide inlets and outlets were punched on PDMS fluidics is limited in the 3D design and world- cured PDMS devices and the PDMS devices were bonded to-chip interface. By using a dual-energy double-sided to cover glasses (0.17 mm-thick) using O2 plasma, com- tape, PMMA and PDMS substrates can be easily and re- pleting the PDMS chip. versibly bonded17,135, enabling broad experimental flex- Second, the PMMA components were fabricated by ibility. cutting the 4-layer design on a 2-mm thick casted PMMA In HMEFC, two PMMA components for world-to- substrate (Kanase, Japan) using a CO2 laser cutter chip interface and electrical application were adhered (VLS3.50, Universal Laser Systems, USA). The layers to a PDMS chip which contained a double-layer mi- were aligned and thermally bonded on a programmable crochannel network where cells were cultured in and ob- automated hot press with temperature and pressure con- served (FIG. 5.b). By double-layer microchannel de- trol (G-12RS2000, Orihara Industrial co., ltd, Japan). sign(FIG. 5.c), experiments with two different cell types Third, dual energy double-sided tape (No.5302A, Nitto, or chemical treatments with four electric field strength Japan) was patterned using the CO2 laser cutter and used to join the PMMA components and the PDMS (EFS) conditions could be performed in a single chip. 17,135 To create multiple dcEFs, the HMEFC was designed chip . A PMMA top reservoir for submerged cell by R-2R resistor ladder configuration17,51,136 to create seeding was also fabricated by four layers of 2 mm-thick theoretical 5.25:2.5:1:0-ratio multiple EFs in section I to PMMA pieces using laser cutting and thermal bonding. IV and section V to VIII (FIG. 5.c). The cells exposed The PMMA top reservoir was affixed on top of the two to the highest dcEF were closest to the outlets to avoid PMMA components through PDMS padding frames with paracrine signaling from electrically-stimulated cells to dual energy double sided tapes, completing the assembly un-stimulated cells. of HMEFC for cell seeding. By using the double-layer design, the hydraulic resis- The detailed description for the HMEFC design, sim- tances in the 10 µm-high, 0.84 mm-wide interconnecting ulation, and fabrication can be found in the Supplemen- channels were much higher than the two 100 µm-high, tary Information. 2 mm-wide main channels where cells resided, limiting the advectional chemical transport and avoiding “cross- contamination” events that further increased the high B. Cell culture and maintenance experimental throughput (see detailed discussion on chip design in Supplementary Information). Glioblastoma cell lines T98G (CRL-1690, ATCC, Coupled numerical simulation of electric field, creeping USA) and U-251MG (IFO50288, JCRB, Japan) were flow, and chemical transport were carried out by finite el- obtained from the respective tissue banks and thawed ement methods (COMSOL Multiphysics 5.3, COMSOL, according to the received instructions. Ethics approval USA). To correctly simulate the system, in-house mea- is not required. T98G and U-251MG were cultured in sured material properties of the minimum essential media minimum essential media α (MEMα) supplemented with −1 ◦ α (MEMα) supplemented with 10% FBS medium were 10% FBS and 2.2 g L under 37 C, 5% CO2 moist at- measured and input in COMSOL. The liquid material mosphere (MCO-18AIC, Sanyo, Japan). The cells were properties of the 3D model was set as water with density subcultured every other day or whenever cells reached of 1002.9 Kg m3, electrical conductivity of 1.536 S m−1, 80% confluency. Mycoplasma contamination was checked dynamic viscosity of 0.946 mPa s, and relative permittiv- every three months using a mycoplasma species-specific bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 9

FIG. 5. The design of hybrid multiple electric field chip (HMEFC). (a) The 3D rendered model of the final HMEFC. Double- layer microchannel design in the PDMS is shown; (b) The schematic diagram of using the HMEFC for electrotaxis experiments. Complex 3D microfluidic structures and world-to-chip interface are established in PMMA components A & B; (c) The channel design. The 10 µm-high first layer structures are shown in cyan.

PCR kit (e-Myco plus, iNtRON, Korea) on a thermocy- inlet/outlet ports. The ECM solutions were allowed to cler (C1000, Bio-Rad, USA). flow into chip passively and incubated for 1 hour at 37◦C. Frozen stocks of the cells were prepared by resuspend- After ECM coatings, the channels were washed once with ing 1×106 log-phase cells in 1 mL CellBanker solution PBS and MEMα. The top reservoir was then filled with (Takara Bio, Japan) and cooled down in a freezing con- MEMα until the inlets and outlets were under the same tainer (Mr. Frosty, Nunc, USA) in -80◦C overnight. The liquid level, balancing the inlet/outlet hydrostatic pres- frozen cells were then transferred into gaseous phase of sure. liquid nitrogen for long term storage (Locator 6, Thermo Log-phase glioblastoma cells were washed with 1X Scientific, USA). PBS, trypsinized (TrypLE, Thermo Fisher Scientific, USA), counted on a benchtop flow cytometer (Muse counting and viability kit, Millipore, USA), centrifuged C. Cell seeding and electrotaxis experiment at 300×g for 5 min, and resuspended in MEMα media with 10% FBS at 106 cells mL−1. Appropriate amount The cell experiment workflow included salt bridge of cell suspension was injected into the microchannels preparation, priming of microchannels, coating substrate using a 200 µL micropipet. Due to the small volume in with ECM, seeding cells, assembly of world-to-chip inter- microchannels, only a minute amount of cell suspension face, and electrotaxis experiment. was needed. The cells were allowed to adhere in the chip ◦ First, sterilized 1% molten agarose (Seakem LE under 37 C, 5% CO2 moist atmosphere for 3 to 5 hours. agarose, Lonza, USA) dissolved in 1X phosphate buffered After cell seeding and adhesion, the top reservoir was saline (PBS) was injected on the salt bridge junctions of removed for optical clarity. A piece of 4 mm-thick PDMS the PMMA component B and allowed to gel (FIG. 5.b). slab punched with two outlet holes (21G, Accu-punch The salt bridge served as a solid electroconductive sep- MP, Syneo Corp, USA) was affixed to the top of compo- aration in electrotaxis experiments between the cell cul- nent B through the first piece of patterned double-sided ture media and electrode, avoiding formation of complex tape (Supplementary FIG. S.15.a), creating an air-tight electrolysis products. seal. Second, 50 µL 99.5% ethanol (Wako, Japan) loaded in To start electrotaxis experiment, fresh media were sup- 200 µL pipet tips were used to wet the microchannels in plied in the reservoirs on component A of HMEFC. Two the PDMS chip by capillary flow and gravity flow17,138. sets of tubings (06419-01, Cole Parmer, USA) with stain- The microchannels were then washed with ultrapure wa- less tubes (21RW, New England Small Tubes, USA) on ter. The inlet/outlet ports were submerged under liquid one ends and double Luer gel dispensing needles on the solutions in all steps afterward to ensure bubble-free mi- opposite ends (23G, Musashi, Japan) were used. The crochannels. tubings were sterilized before priming with 1X PBS with Third, 150 µL of appropriate ECM solutions such as 2.5 mL syringes (Terumo, Japan). The stainless tube GeltrexTM was loaded in tips and inserted on one side of ends were inserted into the 21G holes of the PDMS slab bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 10 on the HMEFC. The syringes were mounted on a syringe D. Microscopy imaging and data processing pump (YSP-202, YMC, Japan) and set in withdrawal mode to perfuse the cells on HMEFC. The HMEFC A Nikon Ti-E automated microscope with Perfect Fo- was completed and ready for electrotaxis experiment cus System and motorized XYZ stage was used to per- (FIG. 5.b). form all microscopy experiments. A 10X phase contrast Two HMEFCs were prepared in one experiment that objective and intermediate magnification of 1.5X were allowed 16 conditions to be screened simultaneously. used for taking 10-minute interval phase contrast images However for data presentation, we only showed the ones with a scientific CMOS (sCMOS) camera with 2×2 bin- with 300 V m−1 and 0 V m−1 dcEF. HMEFCs were af- ning (Orca Flash 4.0, Hamamatsu, Japan) in NIS Ele- fixed in a microscope on-stage incubator (WKSM, Tokai- ment AR software (Nikon, Japan). The spatial resolution hit, Japan). A feedback thin-film K-type thermocouple at this setting was ≈ 0.87 µm pixel−1. All experiments was attached to the glass bottom of a HMEFC (60 µm- were performed in triplicate. thick, Anbesmt, Japan) and regulate the incubator to After each experiment, images were exported from NIS maintain environmental temperature at 37◦C. element software as tiff files and automatically organized A U-shaped PMMA salt bridge filled with 1% molten by XY positions into folders using an in-house devel- agarose in PBS was used to connect the two HMEFCs oped Python script. The cells in the raw images were electrically by inserting into the ajoining PBS reservoirs segmented, tracked, and automatically analyzed using on each HMEFC’s component B. 300 V m−1 dcEFs the Usiigaci software30. Only cells tracked throughout were established in section I and V (FIG. 5.c) through all the frames in one viewfield were analyzed. At least two home-made silver/silver chloride (Ag/AgCl) wire 100 cells in every condition were analyzed for cell-centric electrodes51,139 inserted in the PBS reservoirs on compo- features such as the directedness, speed, and orientation nent B by a source measure unit (2410, Keithley, USA). changes before and after electric field or chemical stimu- Time lapse phase contrast images of each condition were lation (Supplementary TABLE. S.5). taken on an automated phase contrast microscope (Ti- Briefly, the definitions of key cell-centric features used E, Nikon, Japan). The setup is shown in Supplementary to quantify cell electrotaxis (Supplementary FIG. S.11) FIG. S.10. were listed below: For electrotaxis of glioblastoma on different ECM coat- • Directedness ings, different ECM coatings were coated in the microchannels and electrotaxis experiments were performed The directedness of cell electrotaxis was defined as in MEMα with 10% FBS for 6 hours. the average cosine of Euclidean vector and EF vec- N To test if the electrotaxis of gliblastoma cells re- tor, P cos Φi , where Φ was the angle between the 2+ N i quire extracellular Ca , calcium-free Dulbecco’s mini- i=1 mum essential medium (CaFree DMEM, 21068, Gibco, Euclidean vector of each cell migration and the vec- USA) or cation chelators such as ethylenediaminete- tor of applied EF (from anode to cathode), and N traacetic acid (EDTA) (15575, Thermo Fisher Scien- was the total number of analyzed cells. A group of tific, USA) and ethylene glycol-bis(2-aminoethylether)- anodal moving cells held a directedness of -1; and a N,N,N’,N’-tetraacetic acid (EGTA) (08907-84, Nacalai, group of cathodal moving cells held a directedness Japan) were used (Supplementary TABLE. S.3). Before of +1. For a group of randomly migrating cells, the dcEF stimulation, cells were incubated in CaFree DMEM directedness was zero. or MEMα with 10% FBS mixed with calcium chelators at • Speed indicated molar concentrations for 2 hours at 20 µL hr−1. Afterward a direct electric current was applied through The speed of cell electrotaxis was defined as the av- Ag/AgCl wire electrodes in D-PBS buffer on the PMMA erage of cell migration rate to travel the Euclidean dnet N ( )i component B by the source measure unit and cell behav- P telapsed distance, N , where the dnet was the Eu- iors were observed for six hours. i=1 To investigate if voltage-gated ion channels mediate clidean distance traveled by each cell, and telapsed the electrotaxis of glioblastoma cells, the inhibitors were was the the time elapsed, and N was the total num- added to fresh MEMα media with 10% FBS at ap- ber of analyzed cells. propriate working concentrations (Supplementary TA- • Orientation BLE. S.3). The reagents were supplied in the reservoirs on PMMA component A of HMEFCs after cells were The orientation was defined as the average cosine of seeded. The inhibitor-containing media were infused into two times angle between the EF vector and the cell −1 N the channels at 20 µL min for 10 minutes before chang- P cos 2θi −1 long axis, N , where θi was the angle between ing to 20 µL hr and incubated for 30 minutes. A direct i=1 electric current was applied through Ag/AgCl wire elec- the applied EF vector and the long axis of a given trodes in PBS buffer on the PMMA component B by the cell; N is the total number of cells analyzed. A source measure unit and cell behaviors were observed for group of cells aligned perpendicular to the EF held six hours. an orientation of -1; and a group of cells aligned bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 11

in parallel to the applied EF held an orientation 1O. Gallego, “Nonsurgical treatment of recurrent glioblastoma,” of +1. For a group of randomly shaped cells, the Current oncology 22, e273 (2015). 2 average orientation was zero. S. Mallick, R. Benson, A. Hakim, and G. K. Rath, “Manage- ment of glioblastoma after recurrence: A changing paradigm,” The cell-centric features were computed automatically Journal of the Egyptian National Cancer Institute 28, 199–210 and saved as Excel spreadsheets (Office365, Microsoft, (2016). 3A. Hara, T. Kanayama, K. Noguchi, A. Niwa, M. Miyai, USA) as part of the Usiigaci software analysis pipeline. M. Kawaguchi, K. Ishida, Y. Hatano, M. Niwa, and H. Tomita, The data was further statistically inferenced by inputting “Treatment strategies based on histological targets against in- the data in a statistical software (Prism 7, GraphPad vasive and resistant glioblastoma.” Journal of oncology 2019, LLC, USA). All data were presented as the mean ± 95% 2964783 (2019). 4M. Alieva, V. Leidgens, M. J. Riemenschneider, C. A. Klein, confidence interval, which was 1.96 of standard error P. Hau, and J. van Rheenen, “Intravital imaging of glioma bor- of mean, from triplicate experiments. Two-tailed Stu- der morphology reveals distinctive cellular dynamics and contri- dent’s t tests or one-way analyses of variance (ANOVA) bution to tumor cell invasion,” Scientific reports 9, 2054 (2019). with Tukey’s multiple-comparison post-hoc tests were 5A. Dirkse, A. Golebiewska, T. Buder, P. V. Nazarov, A. Muller, performed for statistical testing between two groups or S. Poovathingal, N. H. Brons, S. Leite, N. Sauvageot, D. Sarkis- jan, et al., “Stem cell-associated heterogeneity in glioblastoma multiple groups. The results form one-way ANOVA were results from intrinsic tumor plasticity shaped by the microenvi- reported until noted otherwise. The confidence level to ronment,” Nature communications 10, 1787 (2019). reject a null hypothesis between two datasets was set at 6D. Hambardzumyan and G. Bergers, “Glioblastoma: Defining 95%. A p-value (P, the probability for a true null hypoth- tumor niches.” Trends in cancer 1, 252–265 (2015). 7 esis) less than 0.05 represented a statistical significance H.-F. Tsai, A. Trubelja, A. Q. Shen, and G. Bao, “Tumour-on- a-chip: microfluidic models of tumour morphology, growth and at 95% confidence. microenvironment.” Journal of the Royal Society, Interface 14, 20170137 (2017). 8I. Manini, F. Caponnetto, A. Bartolini, T. Ius, L. Mariuzzi, SUPPLEMENTARY MATERIAL C. Di Loreto, A. P. Beltrami, and D. Cesselli, “Role of microenvironment in glioma invasion: what we learned from in vitro models,” International journal of molecular sciences 19, The supplementary figures, supplementary tables, sup- 147 (2018). plementary videos, and detailed description of HMEFC 9W. Tomaszewski, L. Sanchez-Perez, T. F. Gajewski, and J. H. chip design, simulation, and fabrication are collected in Sampson, “Brain tumor microenvironment and host state: Im- the supplementary material. plications for immunotherapy,” Clinical Cancer Research 25, 4202–4210 (2019). 10M. Diksin, S. J. Smith, and R. Rahman, “The molecular and phenotypic basis of the glioma invasive perivascular niche,” In- DATA AVAILABILITY STATEMENT ternational journal of molecular sciences 18, 2342 (2017). 11A. Hatoum, R. Mohammed, and O. Zakieh, “The unique in- vasiveness of glioblastoma and possible drug targets on extra- The data that support the findings of this study are cellular matrix,” Cancer Management and Research 11, 1843 available from the corresponding author upon reasonable (2019). request. 12C. G. Hales and S. Pockett, “The relationship between local field potentials (lfps) and the electromagnetic fields that give rise to them,” Frontiers in Systems Neuroscience 8, 233 (2014). 13L. Cao, D. Wei, B. Reid, S. Zhao, J. Pu, T. Pan, E. N. Yamoah, ACKNOWLEDGMENTS and M. Zhao, “Endogenous electric currents might guide rostral migration of neuroblasts,” EMBO reports 14, 184–190 (2013). H-.F. Tsai is a Research fellow of Japan Society for the 14J. H. Lee, J. E. Lee, J. Y. Kahng, S. H. Kim, J. S. Park, S. J. Promotion of Science (JSPS Research Fellow) and this Yoon, J.-Y. Um, W. K. Kim, J.-K. Lee, J. Park, et al., “Human glioblastoma arises from subventricular zone cells with low-level work is supported by JSPS KAKENHI (Grant Number driver mutations,” Nature 560, 243–247 (2018). JP1700362). The authors also thank Okinawa Institute 15Y.-J. Huang, G. Hoffmann, B. Wheeler, P. Schiapparelli, of Science and Technology Graduate University (OIST) A. Quinones-Hinojosa, and P. Searson, “Cellular microenvi- for its financial support with subsidy funding from the ronment modulates the galvanotaxis of brain tumor initiating Cabinet Office, Government of Japan. Funders had no cells.” Scientific reports 6, 21583 (2016). 16Y.-J. Huang, P. Schiapparelli, K. Kozielski, J. Green, E. Lavell, role in study design, data collection, the decision to pub- H. Guerrero-Cazares, A. Quinones-Hinojosa, and P. Searson, lish, or preparation of the manuscript. The authors thank “Electrophoresis of cell membrane heparan sulfate regulates gal- Professor Tomoyuki Takahashi (Cellular and Molecular vanotaxis in glial cells.” Journal of cell science 130, 2459–2467 Synaptic Function Unit, OIST) for his invaluable sugges- (2017). 17H.-F. Tsai, K. Toda-Peters, and A. Q. Shen, “Glioblastoma ad- tion and discussion. The authors thank Ms. Yi-Ching hesion in a quick-fit hybrid microdevice,” Biomedical microde- Tsai for her assistance on illustration preparation. vices 21, 30 (2019). H-.F. Tsai and A.Q. Shen declare potential conflict of 18Y.-S. Sun, “Direct-current electric field distribution in the brain interest. A patent application has been submitted by for tumor treating field applications: A simulation study.” Okinawa Institute of Science and Technology Graduate Computational and mathematical methods in medicine 2018, 3829768 (2018). University based on these results with H-.F. Tsai and 19J. G. Lyon, S. L. Carroll, N. Mokarram, and R. V. Bellamkonda, A.Q. Shen as co-inventors. “Electrotaxis of glioblastoma and medulloblastoma spheroidal bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 12

aggregates.” Scientific reports 9, 5309 (2019). matic genomic landscape of glioblastoma.” Cell 155, 462–477 20B. E. Goodman, “Channels active in the excitability of nerves (2013). and skeletal muscles across the neuromuscular junction: basic 37L. N. Kiseleva, A. V. Kartashev, N. L. Vartanyan, A. A. function and pathophysiology,” Advances in physiology educa- Pinevich, and M. P. Samoilovich, “A172 and t98g cell lines tion 32, 127–135 (2008). characteristics,” Cell and Tissue Biology 10, 341–348 (2016). 21W. Taft and R. DeLorenzo, “Regulation of calcium channels 38B. W. Stringer, J. Bunt, B. W. Day, G. Barry, P. R. Jamieson, in brain: implications for the clinical neurosciences.” The Yale K. S. Ensbey, Z. C. Bruce, K. Goasdoué,H. Vidal, S. Charmsaz, journal of biology and medicine 60, 99 (1987). F. M. Smith, L. T. Cooper, M. Piper, A. W. Boyd, and L. J. 22E. M. Kawamoto, C. Vivar, and S. Camandola, “Physiology Richards, “Nuclear factor one b (nfib) encodes a subtype-specific and pathology of calcium signaling in the brain,” Frontiers in tumour suppressor in glioblastoma.” Oncotarget 7, 29306–29320 pharmacology 3, 61 (2012). (2016). 23N. Robil, F. Petel, M.-C. Kilhoffer, and J. Haiech, “Glioblas- 39F. Li, T. Chen, S. Hu, J. Lin, R. Hu, and H. Feng, “Superoxide toma and calcium signaling-analysis of calcium toolbox expres- mediates direct current electric field-induced directional migra- sion,” Int. J. Dev. Biol 59, 407–415 (2015). tion of glioma cells through the activation of akt and erk.” PloS 24F. B. Morrone, M. P. Gehring, and N. F. Nicoletti, “Calcium one 8, e61195 (2013). channels and associated receptors in malignant brain tumor 40R. Fridman, M. C. Kibbey, L. S. Royce, M. Zain, M. Sweeney, therapy.” Molecular pharmacology 90, 403–409 (2016). D. L. Jicha, J. R. Yannelli, G. R. Martin, and H. K. Kleinman, 25J. Pollak, K. G. Rai, C. C. Funk, S. Arora, E. Lee, J. Zhu, N. D. “Enhanced tumor growth of both primary and established hu- Price, P. J. Paddison, J.-M. Ramirez, and R. C. Rostomily, man and murine tumor cells in athymic mice after coinjection “Ion channel expression patterns in glioblastoma stem cells with with matrigel.” Journal of the National Cancer Institute 83, functional and therapeutic implications for malignancy.” PloS 769–774 (1991). one 12, e0172884 (2017). 41R. O. Hynes, “Integrins: versatility, modulation, and signaling 26N. N. Phan, C.-Y. Wang, C.-F. Chen, Z. Sun, M.-D. Lai, and in cell adhesion.” Cell 69, 11–25 (1992). Y.-C. Lin, “Voltage-gated calcium channels: Novel targets for 42M. S. Diamond and T. A. Springer, “The dynamic regulation of cancer therapy.” Oncology letters 14, 2059–2074 (2017). integrin adhesiveness.” Current biology : CB 4, 506–517 (1994). 27A. Maklad, A. Sharma, and I. Azimi, “Calcium signaling in 43S. McLaughlin and M. M. Poo, “The role of electro-osmosis in brain cancers: Roles and therapeutic targeting.” Cancers 11 the electric-field-induced movement of charged macromolecules (2019), 10.3390/cancers11020145. on the surfaces of cells,” Biophysical Journal 34, 85–93 (1981). 28B. Song, Y. Gu, J. Pu, B. Reid, Z. Zhao, and M. Zhao, “Ap- 44F. X. Hart, “Integrins may serve as mechanical transducers for plication of direct current electric fields to cells and tissues in low-frequency electric fields,” Bioelectromagnetics 27, 505–508 vitro and modulation of wound electric field in vivo,” Nature (2006). protocols 2, 1479 (2007). 45K. Zhu, Y. Takada, K. Nakajima, Y. Sun, J. Jiang, Y. Zhang, 29N. Tandon, C. Cannizzaro, P.-H. G. Chao, R. Maidhof, Q. Zeng, Y. Takada, and M. Zhao, “Expression of integrins to A. Marsano, H. T. H. Au, M. Radisic, and G. Vunjak- control migration direction of electrotaxis,” The FASEB Journal Novakovic, “Electrical stimulation systems for cardiac tissue en- , fj–201802657R (2019). gineering,” Nature protocols 4, 155 (2009). 46Y. Taniguchi, H. Ido, N. Sanzen, M. Hayashi, R. Sato- 30H.-F. Tsai, J. Gajda, T. F. Sloan, A. Rares, and A. Q. Shen, Nishiuchi, S. Futaki, and K. Sekiguchi, “The c-terminal region “Usiigaci: Instance-aware cell tracking in stain-free phase con- of laminin beta chains modulates the integrin binding affinities trast microscopy enabled by machine learning,” SoftwareX 9, of laminins.” The Journal of biological chemistry 284, 7820– 230–237 (2019). 7831 (2009). 31L. Kim, Y.-C. Toh, J. Voldman, and H. Yu, “A practical guide 47T. Miyazaki, S. Futaki, H. Suemori, Y. Taniguchi, M. Ya- to microfluidic perfusion culture of adherent mammalian cells.” mada, M. Kawasaki, M. Hayashi, H. Kumagai, N. Nakatsuji, Lab on a chip 7, 681–694 (2007). K. Sekiguchi, and E. Kawase, “Laminin e8 fragments support 32G. A. Villalona, B. Udelsman, D. R. Duncan, E. McGillicuddy, efficient adhesion and expansion of dissociated human pluripo- R. F. Sawh-Martinez, N. Hibino, C. Painter, T. Mirensky, B. Er- tent stem cells.” Nature communications 3, 1236 (2012). ickson, T. Shinoka, et al., “Cell-seeding techniques in vascu- 48J. Cha and P. Kim, “Biomimetic strategies for the glioblastoma lar tissue engineering,” Tissue Engineering Part B: Reviews 16, microenvironment,” Frontiers in Materials 4, 45 (2017). 341–350 (2010). 49L. W. Lau, R. Cua, M. B. Keough, S. Haylock-Jacobs, and 33P. M. Reynolds, C. H. Rasmussen, M. Hansson, M. Dufva, M. O. V. W. Yong, “Pathophysiology of the brain extracellular matrix: Riehle, and N. Gadegaard, “Controlling fluid flow to improve a new target for remyelination,” Nature Reviews Neuroscience cell seeding uniformity,” PloS one 13, e0207211 (2018). 14, 722–729 (2013). 34J. T. Parsons, A. R. Horwitz, and M. A. Schwartz, “Cell ad- 50G. J. Baker, V. N. Yadav, S. Motsch, C. Koschmann, A.-A. Cali- hesion: integrating cytoskeletal dynamics and cellular tension,” nescu, Y. Mineharu, S. I. Camelo-Piragua, D. Orringer, S. Ban- Nature reviews Molecular cell biology 11, 633 (2010). nykh, W. S. Nichols, et al., “Mechanisms of glioma formation: 35C. De Pascalis and S. Etienne-Manneville, “Single and collective iterative perivascular glioma growth and invasion leads to tumor cell migration: the mechanics of adhesions,” Molecular biology progression, vegf-independent vascularization, and resistance to of the cell 28, 1833–1846 (2017). antiangiogenic therapy,” Neoplasia 16, 543–561 (2014). 36C. W. Brennan, R. G. W. Verhaak, A. McKenna, B. Cam- 51H.-F. Tsai, S.-W. Peng, C.-Y. Wu, H.-F. Chang, and J.-Y. pos, H. Noushmehr, S. R. Salama, S. Zheng, D. Chakravarty, Cheng, “Electrotaxis of oral squamous cell carcinoma cells in a J. Z. Sanborn, S. H. Berman, R. Beroukhim, B. Bernard, C.-J. multiple-electric-field chip with uniform flow field.” Biomicroflu- Wu, G. Genovese, I. Shmulevich, J. Barnholtz-Sloan, L. Zou, idics 6, 34116 (2012). R. Vegesna, S. A. Shukla, G. Ciriello, W. K. Yung, W. Zhang, 52A. Soeda, A. Hara, T. Kunisada, S.-i. Yoshimura, T. Iwama, C. Sougnez, T. Mikkelsen, K. Aldape, D. D. Bigner, E. G. and D. M. Park, “The evidence of glioblastoma heterogeneity.” Van Meir, M. Prados, A. Sloan, K. L. Black, J. Eschbacher, Scientific reports 5, 7979 (2015). G. Finocchiaro, W. Friedman, D. W. Andrews, A. Guha, M. Ia- 53Z. Zakaria, A. Tivnan, L. Flanagan, D. W. Murray, M. Salvucci, cocca, B. P. O’Neill, G. Foltz, J. Myers, D. J. Weisenberger, B. W. Stringer, B. W. Day, A. W. Boyd, D. Kögel,M. Rehm, R. Penny, R. Kucherlapati, C. M. Perou, D. N. Hayes, R. Gibbs, D. F. O’Brien, A. T. Byrne, and J. H. M. Prehn, “Patient- M. Marra, G. B. Mills, E. Lander, P. Spellman, R. Wilson, derived glioblastoma cells show significant heterogeneity in C. Sander, J. Weinstein, M. Meyerson, S. Gabriel, P. W. Laird, treatment responses to the inhibitor-of-apoptosis-protein antag- D. Haussler, G. Getz, L. Chin, and T. R. Network, “The so- onist birinapant.” British journal of cancer 114, 188–198 (2016). bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 13

54S. Chen, T. Le, B. A. C. Harley, and P. I. Imoukhuede, “Charac- 74H.-Y. Wang, J.-Y. Li, X. Liu, X.-Y. Yan, W. Wang, F. Wu, terizing glioblastoma heterogeneity via single-cell receptor quan- T.-Y. Liang, F. Yang, H.-M. Hu, H.-X. Mao, Y.-W. Liu, and tification.” Frontiers in bioengineering and biotechnology 6, 92 S.-Z. Zhang, “A three ion channel genes-based signature pre- (2018). dicts prognosis of primary glioblastoma patients and reveals a 55S. L. Perrin, M. S. Samuel, B. Koszyca, M. P. Brown, L. M. chemotherapy sensitive subtype.” Oncotarget 7, 74895–74903 Ebert, M. Oksdath, and G. A. Gomez, “Glioblastoma het- (2016). erogeneity and the tumour microenvironment: implications for 75N. C. K. Valerie, B. Dziegielewska, A. S. Hosing, E. Augustin, preclinical research and development of new treatments.” Bio- L. S. Gray, D. L. Brautigan, J. M. Larner, and J. Dziegielewski, chemical Society transactions 47, 625–638 (2019). “Inhibition of t-type calcium channels disrupts akt signaling and 56N. Schäfer,G. H. Gielen, L. Rauschenbach, S. Kebir, A. Till, promotes apoptosis in glioblastoma cells.” Biochemical pharma- R. Reinartz, M. Simon, P. Niehusmann, C. Kleinschnitz, U. Her- cology 85, 888–897 (2013). rlinger, T. Pietsch, B. Scheffler, and M. Glas, “Longitudinal 76Y. Zhang, N. Cruickshanks, F. Yuan, B. Wang, M. Pahuski, heterogeneity in glioblastoma: moving targets in recurrent ver- J. Wulfkuhle, I. Gallagher, A. F. Koeppel, S. Hatef, C. Papani- sus primary tumors.” Journal of translational medicine 17, 96 colas, J. Lee, E. E. Bar, D. Schiff, S. D. Turner, E. F. Petricoin, (2019). L. S. Gray, and R. Abounader, “Targetable t-type calcium chan- 57E. K. Onuma and S. W. Hui, “A calcium requirement for electric nels drive glioblastoma.” Cancer research 77, 3479–3490 (2017). field-induced cell shape changes and preferential orientation.” 77B. Hille, Ion Channels of Excitable Membranes, Third Edition, Cell calcium 6, 281–292 (1985). 3rd ed. (Sinauer Associates, 2011). 58L. Tung, N. Sliz, and M. R. Mulligan, “Influence of electrical 78G. R. Tibbs, D. J. Posson, and P. A. Goldstein, “Voltage-gated axis of stimulation on excitation of cardiac muscle cells.” Circu- ion channels in the pns: Novel therapies for neuropathic pain?” lation research 69, 722–730 (1991). Trends in pharmacological sciences 37, 522–542 (2016). 59M. Zhao, J. V. Forrester, and C. D. McCaig, “A small, phys- 79M. Strickland, B. Yacoubi-Loueslati, B. Bouhaouala-Zahar, iological electric field orients cell division,” Proceedings of the S. L. F. Pender, and A. Larbi, “Relationships between ion National Academy of Sciences of the United States of America channels, mitochondrial functions and inflammation in human 96, 4942–4946 (1999). aging.” Frontiers in physiology 10, 158 (2019). 60T. A. Banks, P. S. B. Luckman, J. E. Frith, and J. J. Cooper- 80D. R. Trollinger, R. R. Isseroff, and R. Nuccitelli, “Calcium White, “Effects of electric fields on human mesenchymal stem channel blockers inhibit galvanotaxis in human keratinocytes.” cell behaviour and morphology using a novel multichannel de- Journal of cellular physiology 193, 1–9 (2002). vice.” Integrative biology : quantitative biosciences from nano 81M. Aonuma, T. Kadono, and T. Kawano, “Inhibition of an- to macro 7, 693–712 (2015). odic galvanotaxis of green paramecia by t-type calcium channel 61G. H. Markx, “The use of electric fields in tissue engineering: A inhibitors.” Zeitschrift fur Naturforschung. C, Journal of bio- review.” Organogenesis 4, 11–17 (2008). sciences 62, 93–102 (2007). 62L. P. da Silva, S. C. Kundu, R. L. Reis, and V. M. Correlo, 82Y. Atsuta, R. R. Tomizawa, M. Levin, and C. J. Tabin, “L- “Electric phenomenon: A disregarded tool in tissue engineering type voltage-gated ca2+ channel cav1.2 regulates chondroge- and regenerative medicine.” Trends in biotechnology (2019), nesis during limb development,” Proceedings of the National 10.1016/j.tibtech.2019.07.002. Academy of Sciences of the United States of America 116, 63I. Minoura and E. Muto, “Dielectric measurement of individual 21592–21601 (2019). microtubules using the electroorientation method.” Biophysical 83L. A. Pardo, C. Contreras-Jurado, M. Zientkowska, F. Alves, journal 90, 3739–3748 (2006). and W. Stühmer,“Role of voltage-gated potassium channels in 64T. Kim, M.-T. Kao, E. F. Hasselbrink, and E. Meyhöfer,“Ac- cancer,” The Journal of membrane biology 205, 115–124 (2005). tive alignment of microtubules with electric fields.” Nano letters 84H. Wulff, N. A. Castle, and L. A. Pardo, “Voltage-gated potas- 7, 211–217 (2007). sium channels as therapeutic targets,” Nature reviews Drug dis- 65R. Nuccitelli and T. Smart, “Extracellular calcium levels covery 8, 982 (2009). strongly influence neural crest cell galvanotaxis,” The Biolog- 85L. Y. Jan and Y. N. Jan, “Voltage-gated potassium channels and ical Bulletin 176, 130–135 (1989). the diversity of electrical signalling,” The Journal of physiology 66L. J. Shanley, P. Walczysko, M. Bain, D. J. MacEwan, and 590, 2591–2599 (2012). M. Zhao, “Influx of extracellular ca2+ is necessary for electro- 86X. Huang and L. Y. Jan, “Targeting potassium channels in can- taxis in dictyostelium,” J Cell Sci 119, 4741–4748 (2006). cer,” J Cell Biol 206, 151–162 (2014). 67P. Borys, “The role of passive calcium influx through the cell 87A. Grizel, G. Glukhov, and O. Sokolova, “Mechanisms of ac- membrane in galvanotaxis,” Cellular & molecular biology letters tivation of voltage-gated potassium channels,” Acta Naturae 6 18, 187 (2013). (2014). 68L. Guo, C. Xu, D. Li, X. Zheng, J. Tang, J. Bu, H. Sun, Z. Yang, 88N. Comes, A. Serrano-Albarras, J. Capera, C. Serrano-Novillo, W. Sun, and X. Yu, “Calcium ion flow permeates cells through E. Condom, S. R. y Cajal, J. C. Ferreres, and A. Felipe, “In- socs to promote cathode-directed galvanotaxis,” PloS one 10, volvement of potassium channels in the progression of cancer to e0139865 (2015). a more malignant phenotype,” Biochimica et Biophysica Acta 69R. Babona-Pilipos, N. Liu, A. Pritchard-Oh, A. Mok, (BBA)-Biomembranes 1848, 2477–2492 (2015). D. Badawi, M. Popovic, and C. Morshead, “Calcium influx 89D. M. Kim and C. M. Nimigean, “Voltage-gated potassium differentially regulates migration velocity and directedness in re- channels: a structural examination of selectivity and gating,” sponse to electric field application,” Experimental cell research Cold Spring Harbor perspectives in biology 8, a029231 (2016). 368, 202–214 (2018). 90G. Zhang, M. Edmundson, V. Telezhkin, Y. Gu, X. Wei, P. J. 70M. B. Mcferrin and H. Sontheimer, “A role for ion channels in Kemp, and B. Song, “The role of kv1.2 channel in electrotaxis glioma cell invasion,” Neuron glia biology 2, 39–49 (2006). cell migration.” Journal of cellular physiology 231, 1375–1384 71H. Sontheimer, “An unexpected role for ion channels in brain tu- (2016). mor metastasis,” Experimental biology and medicine 233, 779– 91K.-I. Nakajima, K. Zhu, Y.-H. Sun, B. Hegyi, Q. Zeng, C. J. 791 (2008). Murphy, J. V. Small, Y. Chen-Izu, Y. Izumiya, J. M. Penninger, 72R. J. Molenaar, “Ion channels in glioblastoma,” ISRN neurology and M. Zhao, “Kcnj15/kir4.2 couples with polyamines to sense 2011 (2011). weak extracellular electric fields in galvanotaxis.” Nature com- 73R. Wang, C. I. Gurguis, W. Gu, E. A. Ko, I. Lim, H. Bang, munications 6, 8532 (2015). T. Zhou, and J.-H. Ko, “Ion channel gene expression predicts 92J. Morokuma, D. Blackiston, D. S. Adams, G. Seebohm, survival in glioma patients.” Scientific reports 5, 11593 (2015). B. Trimmer, and M. Levin, “Modulation of potassium chan- bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 14

nel function confers a hyperproliferative invasive phenotype on 109E. Chavez-Colorado, Z. Herrera-Carrillo, and J. C. Gomora, embryonic stem cells.” Proceedings of the National Academy “Blocking of t-type calcium channels by tta-a2 reveals a conser- of Sciences of the United States of America 105, 16608–16613 vative binding site for state-dependent antagonists,” Biophysical (2008). Journal 110, 439a–440a (2016). 93M. E. Laniado, S. P. Fraser, and M. B. Djamgoz, “Voltage- 110J. F. Faivre, T. P. Calmels, S. Rouanet, J. L. Javré,B. Cheval, gated k+ channel activity in human prostate cancer cell lines and A. Bril, “Characterisation of kv4.3 in hek293 cells: com- of markedly different metastatic potential: Distinguishing char- parison with the rat ventricular transient outward potassium acteristics of pc-3 and lncap cells,” The Prostate 46, 262–274 current.” Cardiovascular research 41, 188–199 (1999). (2001). 111N. Hatano, S. Ohya, K. Muraki, W. Giles, and Y. Imaizumi, 94E. Venturini, L. Leanza, M. Azzolini, S. Kadow, A. Mattarei, “Dihydropyridine ca2+ channel antagonists and agonists block M. Weller, G. Tabatabai, M. J. Edwards, M. Zoratti, C. Par- kv4.2, kv4.3 and kv1.4 k+ channels expressed in hek293 cells.” adisi, et al., “Targeting the potassium channel kv1. 3 kills British journal of pharmacology 139, 533–544 (2003). glioblastoma cells,” Neurosignals 25, 26–38 (2017). 112T. Kimm and B. P. Bean, “Inhibition of a-type potassium cur- 95S. I. McDonough, L. M. Boland, I. M. Mintz, and B. P. Bean, rent by the peptide toxin snx-482.” The Journal of neuroscience “Interactions among toxins that inhibit n-type and p-type cal- : the official journal of the Society for Neuroscience 34, 9182– cium channels.” The Journal of general physiology 119, 313–328 9189 (2014). (2002). 113H. Vacher, M. Alami, M. Crest, L. D. Possani, P. E. Bougis, and 96A. Lacampagne, F. Gannier, J. Argibay, D. Garnier, and J. Y. M.-F. Martin-Eauclaire, “Expanding the scorpion toxin alpha- Le Guennec, “The stretch-activated ion channel blocker gadolin- ktx 15 family with ammtx3 from androctonus mauretanicus.” ium also blocks l-type calcium channels in isolated ventricu- European journal of biochemistry 269, 6037–6041 (2002). lar myocytes of the guinea-pig.” Biochimica et biophysica acta 114J. K. Maffie, E. Dvoretskova, P. E. Bougis, M.-F. Martin- 1191, 205–208 (1994). Eauclaire, and B. Rudy, “Dipeptidyl-peptidase-like-proteins 97P. Turlapaty, R. Vary, and J. A. Kaplan, “Nicardipine, a new confer high sensitivity to the scorpion toxin ammtx3 to kv4- intravenous calcium antagonist: a review of its pharmacology, mediated a-type k+ channels.” The Journal of physiology 591, pharmacokinetics, and perioperative applications.” Journal of 2419–2427 (2013). cardiothoracic anesthesia 3, 344–355 (1989). 115P. E. Bougis and M.-F. Martin-Eauclaire, “Shal-type (kv4.x) 98J. M. McIntosh, B. M. Olivera, L. J. Cruz, and W. R. Gray, potassium channel pore blockers from scorpion venoms.” Sheng “Gamma-carboxyglutamate in a neuroactive toxin.” The Jour- li xue bao : [Acta physiologica Sinica] 67, 248–254 (2015). nal of biological chemistry 259, 14343–14346 (1984). 116S. Thompson, “Aminopyridine block of transient potassium cur- 99B. M. Olivera, G. P. Miljanich, J. Ramachandran, and M. E. rent.” The Journal of general physiology 80, 1–18 (1982). Adams, “Calcium channel diversity and neurotransmitter re- 117T. Takahashi, “Inward rectification in neonatal rat spinal mo- lease: the omega-conotoxins and omega-agatoxins.” Annual re- toneurones.” The Journal of physiology 423, 47–62 (1990). view of biochemistry 63, 823–867 (1994). 118T. Ishikawa, Y. Nakamura, N. Saitoh, W.-B. Li, S. Iwasaki, and 100P. M. Hinkle, P. A. Kinsella, and K. C. Osterhoudt, “Cadmium T. Takahashi, “Distinct roles of kv1 and kv3 potassium channels uptake and toxicity via voltage-sensitive calcium channels.” The at the calyx of held presynaptic terminal.” The Journal of neu- Journal of biological chemistry 262, 16333–16337 (1987). roscience : the official journal of the Society for Neuroscience 101B. Xu, S. Chen, Y. Luo, Z. Chen, L. Liu, H. Zhou, W. Chen, 23, 10445–10453 (2003). T. Shen, X. Han, L. Chen, and S. Huang, “Calcium signaling 119S. Gründer and X. Chen, “Structure, function, and pharma- is involved in cadmium-induced neuronal apoptosis via induc- cology of acid-sensing ion channels (asics): focus on asic1a,” tion of reactive oxygen species and activation of mapk/mtor International journal of physiology, pathophysiology and phar- network.” PloS one 6, e19052 (2011). macology 2, 73 (2010). 102R. C. Foehring, P. G. Mermelstein, W. J. Song, S. Ulrich, and 120I. Hanukoglu, “Asic and enac type sodium channels: conforma- D. J. Surmeier, “Unique properties of r-type calcium currents tional states and the structures of the ion selectivity filters.” in neocortical and neostriatal neurons.” Journal of neurophysi- The FEBS journal 284, 525–545 (2017). ology 84, 2225–2236 (2000). 121B. K. Berdiev, J. Xia, L. A. McLean, J. M. Markert, G. Y. Gille- 103C. Wormuth, A. Lundt, C. Henseler, R. Müller,K. Broich, spie, T. B. Mapstone, A. P. Naren, B. Jovov, J. K. Bubien, H.-L. A. Papazoglou, and M. Weiergräber, “Review: Cav2.3 r-type Ji, C. M. Fuller, K. L. Kirk, and D. J. Benos, “Acid-sensing voltage-gated ca2+ channels - functional implications in convul- ion channels in malignant gliomas.” The Journal of biological sive and non-convulsive seizure activity,” The open neurology chemistry 278, 15023–15034 (2003). journal 10, 99–126 (2016). 122Y. Tian, P. Bresenitz, A. Reska, L. El Moussaoui, C. P. Beier, 104N. Nejatbakhsh and Z.-p. Feng, “Calcium binding protein- and S. Gründer,“Glioblastoma cancer stem cell lines express mediated regulation of voltage-gated calcium channels linked functional acid sensing ion channels asic1a and asic3.” Scientific to human diseases.” Acta pharmacologica Sinica 32, 741–748 reports 7, 13674 (2017). (2011). 123H.-Y. Yang, R.-P. Charles, E. Hummler, D. L. Baines, and 105M. Ben-Johny and D. T. Yue, “Calmodulin regulation (calmod- R. R. Isseroff, “The epithelial sodium channel mediates the di- ulation) of voltage-gated calcium channels,” The Journal of gen- rectionality of galvanotaxis in human keratinocytes.” Journal of eral physiology 143, 679–692 (2014). cell science 126, 1942–1951 (2013). 106B. A. Simms and G. W. Zamponi, “Neuronal voltage-gated cal- 124D. MBA, M. Mycielska, Z. Madeja, S. P. Fraser, and W. Ko- cium channels: structure, function, and dysfunction,” Neuron rohoda, “Directional movement of rat prostate cancer cells in 82, 24–45 (2014). direct-current electric field: involvement of voltagegated na+ 107C.-M. Tang, F. Presser, and M. Morad, “Amiloride selectively channel activity.” Journal of cell science 114, 2697–2705 (2001). blocks the low threshold (t) calcium channel,” Science 240, 213– 125M. B. A. Djamgoz and R. Onkal, “Persistent current blockers of 215 (1988). voltage-gated sodium channels: a clinical opportunity for con- 108R. L. Kraus, Y. Li, Y. Gregan, A. L. Gotter, V. N. Uebele, S. V. trolling metastatic disease.” Recent patents on anti-cancer drug Fox, S. M. Doran, J. C. Barrow, Z.-Q. Yang, T. S. Reger, K. S. discovery 8, 66–84 (2013). Koblan, and J. J. Renger, “In vitro characterization of t-type 126M. B. A. Djamgoz, S. P. Fraser, and W. J. Brackenbury, “In calcium channel antagonist tta-a2 and in vivo effects on arousal vivo evidence for voltage-gated sodium channel expression in in mice.” The Journal of pharmacology and experimental ther- carcinomas and potentiation of metastasis.” Cancers 11 (2019), apeutics 335, 409–417 (2010). 10.3390/cancers11111675. bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 15

127K.-G. Fischer, N. Jonas, F. Poschenrieder, C. Cohen, M. Kret- nuswami, L. Ni, et al., “Electrical and synaptic integration of zler, S. Greiber, and H. Pavenstädt,“Characterization of a glioma into neural circuits,” Nature 573, 539–545 (2019). na(+)-ca(2+) exchanger in podocytes.” Nephrology, dialysis, 134V. Venkataramani, D. I. Tanev, C. Strahle, A. Studier-Fischer, transplantation : official publication of the European Dialysis L. Fankhauser, T. Kessler, C. Körber, M. Kardorff, M. Ratliff, and Transplant Association - European Renal Association 17, R. Xie, et al., “Glutamatergic synaptic input to glioma cells 1742–1750 (2002). drives brain tumour progression,” Nature 573, 532–538 (2019). 128X.-Q. Dai, A. Ramji, Y. Liu, Q. Li, E. Karpinski, and X.-Z. 135C. F. Carlborg, K. B. Gylfason, A. Ka´zmierczak, F. Dortu, Chen, “Inhibition of trpp3 channel by amiloride and analogs,” M. B. Polo, A. M. Catala, G. M. Kresbach, H. Sohlström, Molecular pharmacology 72, 1576–1585 (2007). T. Moh, L. Vivien, et al., “A packaged optical slot-waveguide 129S. Jeong, S. H. Lee, Y. O. Kim, and M. H. Yoon, “Antinocicep- ring resonator sensor array for multiplex label-free assays in tive effects of amiloride and benzamil in neuropathic pain model labs-on-chips,” Lab on a Chip 10, 281–290 (2010). rats.” Journal of Korean medical science 28, 1238–1243 (2013). 136S. Zhao, K. Zhu, Y. Zhang, Z. Zhu, Z. Xu, M. Zhao, and T. Pan, 130A. Litan and S. A. Langhans, “Cancer as a channelopathy: ion “Electrotaxis-on-a-chip (etc): an integrated quantitative high- channels and pumps in tumor development and progression,” throughput screening platform for electrical field-directed cell Frontiers in cellular neuroscience 9, 86 (2015). migration,” Lab on a Chip 14, 4398–4405 (2014). 131L. Leanza, A. Manago, M. Zoratti, E. Gulbins, and I. Sz- 137Y. Xia and G. M. Whitesides, “Soft lithography,” Angewandte abo, “Pharmacological targeting of ion channels for cancer ther- Chemie International Edition 37, 550–575 (1998). apy: in vivo evidences,” Biochimica et Biophysica Acta (BBA)- 138Y. Wang, D. Lee, L. Zhang, H. Jeon, J. E. Mendoza-Elias, T. A. Molecular Cell Research 1863, 1385–1397 (2016). Harvat, S. Z. Hassan, A. Zhou, D. T. Eddington, and J. Ober- 132A. Müller-Längle, H. Lutz, S. Hehlgans, F. Rödel, K. Rau, holzer, “Systematic prevention of bubble formation and accu- and B. Laube, “Nmda receptor-mediated signaling pathways mulation for long-term culture of pancreatic islet cells in mi- enhance radiation resistance, survival and migration in glioblas- crofluidic device.” Biomedical microdevices 14, 419–426 (2012). toma cells—a potential target for adjuvant radiotherapy,” Can- 139H.-F. Tsai, C.-W. Huang, H.-F. Chang, J. J. Chen, C.-H. Lee, cers 11, 503 (2019). and J.-Y. Cheng, “Evaluation of egfr and rtk signaling in the 133H. S. Venkatesh, W. Morishita, A. C. Geraghty, D. Silverbush, electrotaxis of lung adenocarcinoma cells under direct-current S. M. Gillespie, M. Arzt, L. T. Tam, C. Espenel, A. Pon- electric field stimulation,” PLoS One 8, e73418 (2013). bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Voltage-gated ion channels mediate the electrotaxis of glioblastoma cells in a hybrid PMMA/PDMS microdevice

Hsieh-Fu Tsai1,2∗, Camilo IJspeert1, Amy Q. Shen1∗

1 Micro/Bio/Nanoﬂuidics Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa, Japan 2 Research Fellow of Japan Society for the Promotion of Science ∗ To whom correspondence should be addressed. Email: [email protected]; [email protected]

February 15, 2020

Supplementary Figures

List of Figures

S.1 The complex mechanisms of cell response to electric fields...... 2 S.2 Phase contrast images of T98G & U-251MG cells on various ECMs ...... 3 S.3 The electrotaxis of T98G & U-251MG glioblastoma cells on various ECMs ...... 4 S.4 Time series plot of orientation in electrically stimulated glioblastoma cells...... 4 S.5 Phase contrast images of T98G and U-251MG cells before and after 6 hours 300 V m−1 stimulation. 5 S.6 The electrotaxis of T98G & U-251MG glioblastoma cells in dependence of extracellular calcium 5 S.7 The electrotaxis of T98G & U-251MG glioblastoma cells with inhibition on L-type and N-type VGCCs ...... 6 S.8 The electrotactic speed of T98G & U-251MG glioblastoma cells ...... 7 S.9 Phase contrast images of T98G & U-251MG cells under EF stimulations in different microenvironments after 6 hours...... 8 S.10 The photoimage of the experimental setup for high throughput experiments with the hybrid multiple electric field chip (HMEFC)...... 9 S.11 The single-cell migration parameters extracted from the electrotaxis experiments...... 9 S.12 The circuit models of microfluidic design in the hybrid multiple electric field chip (HMEFC). . 20 S.13 The numerical simulation results of the hybrid multiple electric field chip (HMEFC)...... 22 S.14 Chemical distribution in the hybrid multiple electric field chip (HMEFC) over 10 hours at 20 µL hr−1...... 23 S.15 The assembled hybrid multiple electric field chip (HMEFC)...... 24

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.1: The complex mechanisms of cell response to electric ﬁelds. Adapted from McCaig et al. and Tsai et al1,2.

2 (which wasnotcertifiedbypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmade bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638 available undera CC-BY-NC-ND 4.0Internationallicense ; this versionpostedFebruary20,2020. 3 . The copyrightholderforthispreprint

FIG. S.2: Phase contrast images of T98G & U-251MG cells on various ECMs. Scale bar: 100 µm. bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.3: The electrotaxis of T98G & U-251MG glioblastoma cells on various ECMs after 6 hours under 300 V m−1 EF stimulation. (a) The electrotaxis directedness; (b) The electrotaxis speed; † indicates the electrotaxis groups tested against those without EF stimulation; All other groups are statistically compared to their respective controls on GeltrexTM ECM; n.s. indicates not signiﬁcant; * indicates P < .05; ** indicate P < .01; *** indicate P < .001; **** indicate P < .0001; The numbers in parentheses indicate actual P-values.

FIG. S.4: Time series plot of orientation in electrically stimulated glioblastoma cells. T98G cell results are indicated in closed symbols. U-251MG cell results are indicated in open symbols. Only T98G cells demonstrate prominent perpendicular alignment after electrical stimulation. The dashed lines indicate the 95% conﬁdence interval for respective groups.

4 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.5: Phase contrast images of T98G and U-251MG cells before and after 6 hours 300 V m−1 stimulation. The electric ﬁeld is from left to right. Only T98G cells demonstrate prominent perpendicular allignment after electrical stimulation.

FIG. S.6: The electrotaxis of T98G & U-251MG glioblastoma cells in dependence of extracellular calcium by varying concentration of calcium chelators, EDTA and EGTA. (a) The electrotaxis directedness; (b) The electrotaxis speed; † indicate the electrotaxis groups tested against those without EF stimulation; All groups are statistically compared to controls in cell culture media with 10% FBS; n.s. indicates not signiﬁcant; **** indicate P < .0001; The numbers in parentheses indicate actual P-values.

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.7: The electrotaxis of T98G & U-251MG glioblastoma cells under 300 V m−1 dcEF after 6 hours with pharmacological inhibition on L-type (50 µM Gadolinium [Gd] and Nicardipine) and N-type VGCCs (Conotoxin GVIA). (a) The electrotactic directedness of the two cells. (b) The electrotactic speed of the two cells. † indicate the electrotaxis group tested against those with cytochrome C which prevents adsorption of short peptides to experimental apparatus; All other groups are statistically compared to their respective controls in cell culture media with 10% FBS; n.s. indicates not signiﬁcant; **** indicate P < .0001.

6 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.8: The electrotactic speed of T98G & U-251MG glioblastoma cells under 300 V m−1 dcEF after 6 hours with pharmacological inhibition on various ion channels. † indicates the electrotaxis tested against those without EF stimulation; ‡ indicate the electrotaxis group tested against those with cytochrome C which prevents adsorption of short peptides to experimental apparatus; All other groups are statistically compared to their respective controls in cell culture media with 10% FBS; n.s. indicates not signiﬁcant; ** indicate P < .01; *** indicate P < .001; **** indicate P < .0001; The numbers in parentheses indicate actual P-values.

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.9: Phase contrast images of T98G & U-251MG cells under EF stimulations in diﬀerent microenvironments after 6 hours. The scale bars represent 100 µm. 4-AP: 4-aminopyridine. T98G cells demonstrate perpendicular alignment after electric ﬁeld stimulation in cell culture media with or without FBS. However, without calcium and FBS in the cell culture medium, T98G cells’ viability decreases. Furthermore, in cell culture medium supplemented with 4 mM 4-AP, T98G also lose viability and start to detach from the substrate. These trends were not observed in the U-251MG cells, suggesting heterogeneity between cell lines.

8 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.10: The photoimage of the experimental setup for high throughput experiments with the hybrid multiple electric ﬁeld chip (HMEFC). The red dashed box indicates the magniﬁed region with the hybrid PMMA/PDMS HMEFC in an on-stage incubator. SMU: source measure unit

FIG. S.11: The single-cell migration parameters extracted from the electrotaxis experiments.

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder,Supplementary who has granted bioRxiv Tables a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. List of Tables

S.1 Commonly used extracellular matrix coatings for cell migration...... 10 S.2 The results of glioblastoma electrotaxis on various ECM coatings in HMEFC after 6 hours. . . 11 S.3 Pharmacological inhibitors to study various ion channels in glioma electrotaxis...... 12 S.4 The detailed results of glioblastoma electrotaxis in the presence of antagonists against ion channels or calcium signaling after 6 hours...... 13 S.5 Step-centric and cell-centric variables for describing single-cell migration...... 14 S.6 Physical characteristics of the double-layer hybrid multiple electric ﬁeld chip (HMEFC). . . . . 19

TABLE. S.1: Commonly used extracellular matrix coatings for cell migration.

ECM Stock conc. Working conc. Common use

Positively charged synthesized amino poly(D-lysine) 500 µg mL−1 50 µg mL−1 acids for general cell adhesion.

General coating for cell attachment, Collagen I 3 mg mL−1 50 µg mL−1 migration and morphogenesis

General coating for cell attachment, Collagen IV 3 mg mL−1 50 µg mL−1 migration and morphogenesis

0.01% Positively charged synthesized amino poly(L-ornithine) 100 µg mL−1 (0.1 mg mL−1) acids for general cell adhesion.

poly(L-ornithine) 100 µg mL−1+ 50 µg mL−1 General coating for cells from n/a +Laminin 20 µg mL−1+ 10 µg mL−1 central or peripheral nervous system.

50 µg mL−1 Maintenance of ESCs; Laminin 100 µg mL−1 10 µg mL−1 Epithelial diﬀerentiation; Neurite outgrowth

Feeder-free maintenance of stem cells Vitronectin 0.5 mg mL−1 5 µg mL−1 and primary neurons.

Fibrillar matrices for stem cells Fibronectin 1 mg mL−1 10 µg mL−1 and neural outgrowth.

iMatrix 511 Truncated form of laminin for enhanced 0.5 mg mL−1 50 µg mL−1 (Laminin 511-E8) iPSC feeder-free culture

0.12 – 0.18 mg mL−1 0.12 – 0.18 mg mL−1 Maintenance of ESCs; Geltrex (ready-to-use) (ready-to-use) Endothelial capillary formation

10 (which wasnotcertifiedbypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmade bioRxiv preprint

TABLE. S.2: The results of glioblastoma electrotaxis on various ECM coatings in HMEFC after 6 hours. sem: standard error of mean. doi: Concentration EF T98G U-251MG

ECM https://doi.org/10.1101/2020.02.14.948638 (µg mL−1) (V m−1) Speed Speed N Directedness sem sem Orientation sem N Directedness sem sem Orientation sem (µm hr−1) (µm hr−1) 120–180 300 2634 -0.66 0.02 9.52 0.33 -0.75 0.03 276 0.47 0.06 6.66 0.50 -0.07 0.07 Geltrex (ready-to-use) 0 2014 0.15 0.05 4.62 0.32 0.04 0.05 295 -0.08 0.07 3.97 0.41 0.04 0.07 300 1047 -0.61 0.03 2.95 0.13 -0.17 0.04 470 0.03 0.03 6.79 0.22 -0.10 0.04 Poly(D-lysine) 50 0 1164 0.00 0.03 1.98 0.09 -0.10 0.03 370 0.03 0.04 5.19 0.21 -0.04 0.04 available undera 300 960 -0.43 0.03 6.01 0.21 -0.52 0.03 278 0.14 0.02 15.50 0.30 -0.11 0.04 Poly(L-ornithine) 100 0 1059 0.06 0.03 3.81 0.17 -0.04 0.04 189 0.42 0.04 7.70 0.29 0.03 0.06 300 221 -0.38 0.09 8.04 0.71 -0.15 0.10 233 0.23 0.07 3.97 0.35 -0.11 0.08 Poly(L-ornithine) 100/50 0 320 0.01 0.07 6.38 0.57 -0.04 0.07 265 0.18 0.06 3.05 0.28 -0.07 0.07 CC-BY-NC-ND 4.0Internationallicense +

300 574 -0.44 0.06 8.90 0.75 -0.22 0.07 254 0.39 0.06 3.69 0.31 -0.12 0.06 ; Laminin 20/10 this versionpostedFebruary20,2020. 11 0 665 -0.26 0.07 8.22 0.76 0.06 0.07 226 0.08 0.04 3.59 0.24 -0.03 0.05 300 1474 -0.83 0.01 11.94 0.30 -0.69 0.02 194 0.31 0.04 9.37 0.27 -0.00 0.07 50 0 1059 -0.69 0.02 7.25 0.24 -0.02 0.04 186 0.29 0.05 5.65 0.34 -0.02 0.08 Laminin 300 293 0.34 0.04 3.43 0.19 -0.28 0.04 456 -0.17 0.05 6.64 0.37 -0.10 0.05 10 0 404 0.34 0.03 3.71 0.18 0.09 0.04 302 -0.02 0.06 4.99 0.35 0.08 0.06 300 359 0.16 0.04 3.14 0.15 -0.15 0.04 166 0.50 0.05 7.11 0.52 0.07 0.06 Collagen I 50 0 891 0.11 0.04 4.93 0.30 -0.05 0.04 135 -0.03 0.12 6.08 0.92 0.04 0.13 300 424 -0.13 0.03 4.39 0.21 -0.29 0.03 207 -0.40 0.04 8.87 0.56 -0.08 0.05 Collagen IV 50 0 352 -0.08 0.04 2.65 0.13 0.01 0.04 73 0.03 0.08 6.58 0.70 0.05 0.09 .

300 613 -0.80 0.01 4.46 0.10 -0.20 0.03 184 0.01 0.05 4.76 0.33 -0.15 0.05 The copyrightholderforthispreprint Vitronectin 5 0 342 0.15 0.04 2.56 0.16 -0.05 0.04 200 -0.04 0.05 4.91 0.32 0.14 0.05 300 601 -0.34 0.03 7.98 0.28 -0.39 0.03 163 0.10 0.12 15.45 1.75 -0.31 0.13 iMatrix-511 50 0 385 -0.09 0.04 7.30 0.28 -0.06 0.04 122 -0.10 0.09 14.79 1.21 -0.07 0.10 300 553 -0.50 0.02 6.01 0.24 -0.55 0.03 263 0.14 0.07 5.71 0.43 -0.12 0.07 Fibronectin 10 0 543 -0.25 0.03 5.10 0.19 -0.11 0.03 658 -0.02 0.04 4.93 0.26 -0.02 0.04 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

TABLE. S.3: Pharmacological inhibitors to study various ion channels in glioma electrotaxis.

Inhibitors Stock conc. Working conc. Actions

50 µM Cadmium chloride 1 M HVA VGCC inhibitor 100 µM

Gadolinium chloride 1 M 50 µM HVA L-type VGCC inhibitor

100 µM Nickel chloride 1 M LVA T-type VGCC inhibitor 500 µM

EDTA 0.5 M 2 mM Cation chelators (e.g., Ca2+)

Cation chelators EGTA 0.5 M 5 mM (more selective toward Ca2+)

Nicardipine 20 mM DMSO 10 µM L-type VGCC inhibitor

ω-Conotoxin GVIA 0.1 mM 3 µM N-type VGCC inhibitor

ω-Agatoxin IVA 0.1 mM 200 nM P/Q-type VGCC inhibitor

ω-Conotoxin MVIIC 0.2 mM 500 nM P/Q-type VGCC inhibitor

R-type VGCC inhibitor; ω-Theraphotoxin-Hg1a (SNX482) 0.1 mM 500 nM A-type K+ channel inhibitor

TTA-A2 0.1 M DMSO 100 µM T-type VGCC inhibitor

T-type VGCC inhibitor; Amiloride hydrochloride 5 mM 100 µM ENaC/ASIC/Degenerin inhibitor

Calmidazolium 10 mM DMSO 100 nM Calmodulin antagonist

AmmTx3 0.1 mM 5 µM A-Type K+ channel inhibitor

1 mM 4-aminopyridine 50 mM A-Type K+ channel inhibitor 4 mM

Na+/Ca2+ exchanger inhibitor;

Benzamil hydrochloride 10 mM DMSO 5 µM ASIC/ENaC inhibitor;

TRPP inhibitor; TRPA1 inhibitor

12 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

TABLE. S.4: The detailed results of glioblastoma electrotaxis in the presence of antagonists against ion channels or calcium signaling after 6 hours. CytoC: Cytochrome C; sem: standard error of mean.

CytoC EF T98G U-251MG Condition (1 mg mL−1) (V m−1) Speed Speed N Directedness sem sem N Directedness sem sem (µm hr−1) (µm hr−1) 300 128 -0.69 0.04 9.99 0.57 288 0.58 0.03 7.33 0.39 FBS no 0 264 -0.18 0.04 4.85 0.21 330 0.08 0.04 4.98 0.25 300 212 -0.46 0.04 5.50 0.35 250 0.16 0.05 3.27 0.23 Serum free no 0 338 -0.11 0.04 5.16 0.35 186 0.26 0.05 3.67 0.25 CaFree DMEM 300 116 -0.80 0.03 9.95 0.70 104 0.44 0.06 8.15 1.14 no (10% FBS) 0 117 -0.22 0.06 8.07 0.58 108 -0.22 0.07 7.19 0.61 CaFree DMEM 300 0 n/a n/a n/a n/a 119 0.21 0.07 9.02 1.09 no (Serum free) 0 106 0.24 0.07 6.56 0.73 123 -0.03 0.07 9.24 0.79 Cytochrome C 300 476 -0.51 0.03 9.34 0.37 302 0.58 0.03 4.73 0.28 yes (1 mg ml−1) 0 334 -0.21 0.04 5.61 0.27 368 0.18 0.04 4.22 0.21 Cd 300 119 -0.21 0.07 6.71 0.62 314 0.48 0.03 4.92 0.20 no (50 µM) 0 196 -0.14 0.05 4.11 0.37 276 0.02 0.03 5.12 0.21 Cd 300 339 -0.25 0.04 4.64 0.22 232 0.34 0.04 3.37 0.18 no (100 µM) 0 294 -0.12 0.04 2.88 0.17 348 -0.21 0.04 2.21 0.11 Gd 300 118 -0.67 -0.27 10.82 0.65 188 0.59 0.04 5.56 0.37 no (50 µM) 0 209 0.05 0.04 7.21 0.59 340 -0.11 0.05 5.47 0.23 Ni 300 359 -0.81 0.02 13.97 0.46 485 0.21 0.03 2.94 0.17 no (100 µM) 0 400 -0.13 0.04 5.33 0.22 423 -0.09 0.03 5.50 0.29 Ni 300 226 -0.55 0.04 10.64 0.55 350 0.25 0.04 4.48 0.25 no (500 µM) 0 264 -0.35 0.04 6.15 0.37 380 0.23 0.04 3.33 0.19 EDTA 300 461 -0.54 0.03 8.62 0.31 359 0.19 0.04 7.25 0.35 no (2 mM) 0 473 -0.16 0.03 7.13 0.25 524 0.06 0.03 5.85 0.25 EGTA 300 141 -0.57 0.07 7.22 1.71 107 0.48 0.06 7.74 2.09 no (1 mM) 0 153 -0.07 0.07 7.15 1.61 109 0.01 0.08 2.49 1.95 EGTA 300 1028 -0.68 0.02 6.97 0.18 587 0.14 0.03 4.38 0.23 no (2 mM) 0 1367 -0.20 0.02 2.93 0.08 365 0.04 0.04 4.49 0.24 EGTA 300 160 -0.36 0.05 2.21 0.11 472 -0.14 0.03 2.42 0.09 no (5 mM) 0 139 0.18 0.07 1.77 0.15 376 -0.04 0.03 2.42 0.11 Nicardipine 300 218 -0.65 0.04 8.19 0.43 315 0.64 0.03 6.94 0.35 no (10 µM) 0 128 0.04 0.06 6.92 0.50 297 -0.03 0.04 5.43 0.31 ω-Conotoxin GVIA 300 211 -0.67 0.03 12.43 0.63 320 0.44 0.03 6.01 0.33 yes (3 µM) 0 108 0.14 0.07 4.98 0.44 194 0.01 0.05 3.94 0.29 ω-Agatoxin IVA 300 145 -0.53 0.05 9.64 0.61 177 0.13 0.05 7.13 0.47 yes (200 nM) 0 105 -0.08 0.07 4.77 0.47 235 -0.14 0.05 6.79 0.38 ω-Conotoxin MVIIC 300 309 -0.75 0.02 12.59 0.51 477 -0.09 0.03 4.61 0.22 yes (500 nM) 0 362 -0.30 0.03 6.55 0.25 375 -0.24 0.03 2.77 0.15 ω-Theraphotoxin-Hg1a 300 153 -0.27 0.06 7.56 0.48 341 0.41 0.03 6.31 0.34 yes (500 nM) 0 127 0.01 0.08 5.24 0.54 369 -0.12 0.04 4.99 0.26 TTA-A2 300 677 -0.69 0.02 6.65 0.17 625 0.47 0.03 6.00 0.27 no (200 µM) 0 501 -0.20 0.03 7.73 0.23 556 0.11 0.03 4.53 0.20 Continue on next page

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Continued from previous page CytoC EF T98G U-251MG Condition (1 mg mL−1) (V m−1) Speed Speed N Directedness sem sem N Directedness sem sem (µm hr−1) (µm hr−1) Amiloride hydrochloride 300 352 -0.77 0.02 12.96 0.42 332 0.32 0.04 4.87 0.30 no (100 µM) 0 524 0.06 0.03 6.02 0.23 413 -0.19 0.04 3.57 0.20 Calmidazolium 300 229 -0.80 0.02 12.26 0.45 346 0.22 0.04 6.42 0.32 no (100 nM) 0 437 -0.18 0.03 5.75 0.24 449 -0.01 0.03 4.25 0.22 AmmTx3 300 549 -0.39 0.03 4.42 0.17 816 0.13 0.02 6.06 0.15 yes (5 µM) 0 337 -0.10 0.04 3.63 0.20 555 -0.15 0.03 4.27 0.16 4-aminopyridine 300 1335 0.11 0.02 4.95 0.11 1033 0.07 0.02 3.37 0.10 no (1 mM) 0 1220 0.56 0.02 4.76 0.10 570 0.08 0.03 3.66 0.16 4-aminopyridine 300 963 -0.79 0.01 4.89 0.09 749 0.11 0.03 2.83 0.11 no (4 mM) 0 678 0.85 0.01 3.69 0.06 654 0.06 0.03 2.48 0.10 Benzamil hydrochloride 300 856 -0.42 0.02 5.48 0.14 428 0.29 0.03 6.05 0.26 no (5 µM) 0 1108 -0.08 0.02 4.43 0.11 455 0.14 0.03 6.81 0.27

TABLE. S.5: Step-centric and cell-centric variables for describing single-cell migration.

Step-centric features

p 2 2 Instantaneous displacement di = (xi − xi−1) + (yi − yi−1)

Instantaneous speed si = d(pi−1, pi)/∆t

−1 (yi−yi−1) Turning angle αi = tan (xi−xi−1)

Directional autocorrelation dir − auti = cos (αi − αi−1) N N N cos Φ P instanti P (xi−xi−1) 1 P (xi−xi−1) Instantaneous directedness Indexdirectedness = = = √ instant N di×N N 2 2 i=1 i=1 i=1 (xi−xi−1) +(yi−yi−1) Cell-centric features N P Cumulative distance dtotal = d(pi−1, pi) i=1 Net trigonometric distance p 2 2 dnet = d(p0, pN ) = (xN − x0) + (yN − y0) (Euclidean distance)

dnet Euclidean velocityv ¯net = telapsed End-point directionality ratio ep dr = dnet dtot N P cos 2θi Orientation Indexorientation = N i=1 N N N P cos Φi P (xi−x0) 1 P (xi−x0) Directedness Indexdirectedness = = = √ N dnet×N N 2 2 i=1 i=1 i=1 (xi−x0) +(yi−y0)

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder,Supplementary who has granted VideosbioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

VIDEO. S.1: Video clip showing the electrotaxis of T98G glioblastoma cells under 300 V m−1 dcEF for six hours and the respective tracking results using Usiigaci software.

VIDEO. S.2: Video clip showing the random migration of T98G glioblastoma cells for six hours and the respective tracking results using Usiigaci software.

15 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

VIDEO. S.3: Video clip showing the electrotaxis of U-251MG glioblastoma cells under 300 V m−1 dcEF for six hours and the respective tracking results using Usiigaci software.

VIDEO. S.4: Video clip showing the random migration of U-251MG glioblastoma cells for six hours and the respective tracking results using Usiigaci software.

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

VIDEO. S.5: Video clip showing the electrotaxis of T98G glioblastoma cells suppressed with 500 nM SNX482 on R-type VGCC under 300 V m−1 dcEF for six hours and the respective tracking results using Usiigaci software.

VIDEO. S.6: Video clip showing the electrotaxis of U-251MG glioblastoma cells suppressed with 200 nM Agatoxin IVA on P/Q-type VGCC under 300 V m−1 dcEF for six hours and the respective tracking results using Usiigaci software.

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

VIDEO. S.7: Video clip showing the electrotaxis of T98G glioblastoma cells suppressed with 4-AP on A-type VGKC under 300 V m−1 dcEF for six hours and the respective tracking results using Usiigaci software.

VIDEO. S.8: Video clip showing the electrotaxis of U-251MG glioblastoma cells suppressed with 4-AP on A-type VGKC under 300 V m−1 dcEF for six hours and the respective tracking results using Usiigaci software.

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) isSupplementary the author/funder, who has granted Information bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 Hybrid multiple electric ﬁeld chip (HMEFC) design, simulation and fabrication

1.1 HMEFC design

The HMEFC was designed by the hybrid PMMA/PDMS approach3 (FIG. 5), in which the prototyping disadvantages in both materials could be mitigated while the advantages can be combined. In PMMA, complex

3D structures for fluidic routing or reservoir and world-to-chip interface could be quickly prototyped by CO2 laser cutting and thermal bonding. However, spatial resolution using this approach was not high to create reliable microfluidic environments. In contrast, highly precise quasi-two dimensional microstructures could be fabricated using the soft lithography technique in PDMS microfluidic chip, however, standard soft lithogarphy was limited by 3D design complexity and world-to-chip interface. Using a dual-energy double-sided tape, PMMA and PDMS could be easily and reversibly bonded3,4, enabling broad flexibility in microfluidic design and experiments. In HMEFC, two PMMA components for world-to-chip interface and electrical application were adhered to a PDMS chip which contained a double-layer microchannel where cells were cultured in and observed (FIG. 5b). The electric field was applied from agrose interface next to the outlets that were connected to syringes. By the double-layer microchannel design, four different strengths of direct current electric field (dcEF) were created in two main channels in a HMEFC and the chemical transport across them were limited by manipulating hydraulic resistances (FIG. 5.c). Two different cell types or chemical treatments could be used in the main channels. The microchannel characteristics were shown in TABLE. S.6. Two 43.8 mm-long, 2 mm-wide, 100 µm-high main channels were connected with three 1.5 mm long, 0.84 mm wide, 10 µm-high interconnecting channels at a spacing of 8.95 mm. The electrical equivalent circuit and hydraulic equivalent circuit were modeled as shown in FIG. S.12.

TABLE. S.6: Physical characteristics of the double-layer hybrid multiple electric ﬁeld chip (HMEFC).

Double layer Channel type Interconnection Main Channel dimension (L×W×H) (mm3) 1.5×0.84×0.01 8.95×2×0.1 2wh 0.01976 0.1905 Hydraulic diameter ( w+h ) (mm) −1 Relative RΩ (mm ) 178.99 44.75 −3 9 Rhydraulic (Pa s m )(×10 ) 21616.4 55.2

Relative Rhydraulic 391.88 1 Flow velocity simulated by PSPICE at 20 µL min−1 (m s−1) 2.02E-21 1.75E-3 Reynolds number 4.01E-20 0.333 P´ecletnumber 8.97E-16 7460

To create multiple dcEFs, the HMEFS were designed by R-2R resistor ladder configuration3,5,6 creating different electric fields in section I to IV and section V to VIII (FIG. S.12.a). The cells exposed to the highest dcEF were nearest to the outlets to avoid paracrine signaling from electrically-stimulated cells to un-stimulated cells. According to Ohm’s law, the electrical resistance of a resistor, R, was proportional to the length and inversely proportional to the cross-sectional area: l ρl R = ρ = , (1) A wh where ρ, l, A, w, and h were the electrical resistivity of the medium, the length, the cross-sectional area, the

width, and the height of the microchannel, respectively. Assuming the electrical resistance of R1 being r, the

19 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.12: The circuit models of microfluidic design in the hybrid multiple electric field chip (HMEFC). (a) The electric field equivalent circuit; (b) The hydraulic equivalent circuit. The italic r depicts relative electrical resistance or relative hydraulic resistance in each circuit component.

relative electrical resistances of other segments (R2 :R11) could be calculated accordingly. The electric current ﬂowing through each resistor was calculated by Kirchhoﬀ’s circuit law and simulated using PSPICE in the electronic design automation software (OrCAD Lite, Cadence Design Systems, USA) by:

10r × i2 − 4r × i1 − 4r × i3 = 0, (2a)

10r × i3 − 4r × i2 = 0. (2b)

By solving the system of equations in Equation 2b, the ratio of electric currents between each segment, hence the ratio of electric ﬁeld strengths, was derived in Equation 3c:

i1 : i2 : i3 = EI :EII :EIII :EIV (3a)

= EV :EVI :EVII :EVIII (3b) = 5.25 : 2.5:1:0 (3c)

The similar analysis could be performed on the hydraulic equivalent circuit of HMEFC using the electrical- hydraulic analogy7 (FIG. S.12.b). The interconnecting channels and main channels had width much bigger than height, thus, the hydraulic resistance can be calculated using the below equation8:

12µl R = , (4) H 3 h wh (1 − 0.53 w )

where µ is the fluid viscosity, l is the channel length, w is the channel width, and h is the channel height. The hydraulic resistances in the 10 µm-high interconnecting channel were much higher than the two 100 µm- high, 2 mm-wide main channels were cells resided, limiting the advectional chemical transport. The flow rates in interconnecting channels and main channels differed by 18 orders as simulated by PSPICE analysis (TABLE. S.6). Two dimensionless numbers such as Reynolds number and Pécletnumber can be used to characterize the fluid flow and chemical transport in the microfluidic system. The Reynolds number, <, is a dimensionless parameter to determine if the system is in the laminar or turbulent regime9. It is a measure of competition between the inertia force over the viscous force in the flow. In general, most microfluidic chips for cell studies involve Reynolds numbers much smaller than 1 and the flows in the lab-on-chip devices are observed to be laminar.

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made ρUL available under aCC-BY-NC-ND< = c 4.0, International license. (5) µ

where Lc is a characteristic length scale related to the flow field, U is the characteristic velocity, ρ is the density, and µ is the dynamic viscosity of the fluid. In HMEFC, the characteristic length was the hydrodynamic radius 2wh ( w+h ), where w and h were width and height of microchannels. The Pécletnumber describes the proportional relationship of chemical transport between convective fluxes and diffusive fluxes. In a system of high Pécletnumber, the diffusion is negligible whereas in low Pécletnumber system, the scalar (i.e., solute concentration) distribution governed by diffusion follows the Fick’s Law9:

UL P = c , (6) e D

where U is the characteristic velocity, Lc is the characteristic length, and D is the diffusion coefficient of the scalar. In HMEFC, the Reynolds number and Pécletnumber were much lower in the interconnecting channels compared to those of the main channels, suggesting that the chemical transport through the interconnecting channels that would cause “cross-contamination” event was diffusion-limited.

1.2 HMEFC simulation

The 3D design model HMEFC was exported from AutoCAD and imported into COMSOL Multiphysics 5.3 (COMSOL Inc, USA). Steady-state coupled simulations using three modules, creeping flow, chemical transport of diluted species, and electric currents, were performed. To correctly simulate the system, in-house measured material properties of the minimum essential media α (MEMα) for cell culture supplemented with 10% FBS medium were input. The material in the 3D model was set as water with density of 1002.9 Kg m3, electrical conductivity of 1.536 S m−1, dynamic shear viscosity of 0.946 mPa s, and relative permittivity of 80. The numerically simulated electric field ratio was 4.99: 2.45:1:0 in section I to IV and section V to VIII with limited chemical transport across the interconnecting channels as designed. The boundary conditions were input accordingly for creeping flow, electric current, and chemical transport of diluted species as shown in FIG. 5. For creeping flow, 20 µL min−1 flow rate was used. For chemical transport, 100 mol mm−3 of 40 kDa dextran was set to inflow in the top main channel while no dextran was flew in the bottom channel. The 40 kDa had a diffusion coefficient of 44.7 µm2 s−110. For establishment of 300 V m−1 electric field, 485.5 µA m−2 was set at the anode and electric potential of 0 V was set at the cathode. The dcEFs in section I to IV and V to VIII were simulated as 4.99:2.45:1:0 (FIG. S.13.a) which were comparable to the theoretical calculation. The chemical distribution of the 40 kDa dextran was shown in FIG. S.13.b. By adjusting the scale of the colormap, the “cross-contamination” events where chemical leaches from one main channel to another through the interconnecting channels could be seen more clearly (FIG. S.13.c). In the double-layer design, it was more diffusion-dominant and the cross-contamination was limited. These results shows the HMEFCs could create multiple dcEFs without cross-contamination of chemicals, increasing the experimental throughput.

1.3 Diﬀusion-limited chemical transport validation in HMEFC

The diffusion-limited chemical transport in HMEFC was visualized by flowing 40 kDa fluorescein-dextran (FD40, Sigma-Aldrich, USA) with U-251MG cells or 40 kDa tetramethylrhodamine-dextran (D1842, Thermo Fisher Scientific, USA) in Fluorobrite DMEM (Gibco, USA). The dyed media were loaded separately in two 2.5 mL syringes (Terumo, Japan) and the channels was primed at 20 µL min−1 for 10 minutes before reducing to 20 µL hr−1 (cell culture experiment flow rate), followed by capturing microscopy images with 10 minute time lapse epifluorescence microscopy over ten hours (Ti-E, Nikon, Japan). As shown in FIG. S.14, the chemical transport was stable over the course of 10 hours at cell culture-relevant

21 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.13: The numerical simulation results of the hybrid multiple electric field chip (HMEFC). (a) The electric field distribution in HMEFC. Application of a direct current electric field from the two outlets (nearest to section I) creates 4.99:2.45:1:0 stepwise electric fields for high throughout electrotaxis study; (b) The coupled chemical transport (convection and electrokinesis) simulated in HMEFC. 100 mol m−3 of 40 kDa dextran was flowing in the top channel and 0 mol m−3 was flowing in the bottom channel; (c) The coupled chemical transport simulated in HMEFC with the color scaled adjusted to show the “cross-contamination” event. The red dashed box shows the magnified region of an interconnecting channel. Only a minute amount of chemical diffuse through the interconnecting channel.

flow rates. Furthermore, when cells were injected in the top channel of the double-layer microfluidic chip, the cells were not able to pass through 10 µm-height interconnecting channels and were restricted to its main channel. When different cell types were seeded in separate channels, the double-layer microchannel design could further increase the experimental throughput.

1.4 HMEFC fabrication

A HMEFC was composed of a PMMA top reservoir, two PMMA components, and a PDMS chip (FIG. S.15). For high throughput experiments, another U-shaped PMMA agarose salt bridge made electrical conduction between two HMEFCs.

PDMS chip fabrication In the double-layer HMEFC, the first layer was 10 µm-high and the second layer was 100 µm. The design was designed in AutoCAD and exported to KLayout. Several cross-shaped alignment markers were included in the design to assist alignment during making of the mold for double-layer microfluidic chip. A 10 µm-thick layer of photoresist (DWL-5, Micro Resist Technology, Germany) was spin-coated on a 100 mm silicon wafer and soft baked. The first design layer was directly written by the maskless lithographic writer (80 mJ cm−2, DL-1000, Nano Systems Solutions, Japan) and subsequently hard-baked. Next, without development, a layer of 100µm-thick photoresist (SU-8 3050, MicroChem Corp, USA) was spin-coated on the wafer and soft baked. A chrome mask (CBL5006Du-AZPFS, Clean Surface Technology, Japan) for the second layer was fabricated using the maskless lithographic writer and the pattern was etched away by etchant (651826, Sigma Aldrich, USA). Using the alignment markers, the silicon wafer with the first-layer structures was aligned to the second layer structures on the chrome mask on the mask aligner (MA/BA6, SUSS MicroTec, Germany) and subsequently exposed (45 s of i-line UV irradiation). The unexposed photoresist was dissolved away in propylene glycol methyl ether acetate (PGMEA, Sigma Aldrich, USA) and the wafer was washed thoroughly with isopropanol, water, and dried with nitrogen gas. The wafer with microstructures was passivated by 1H, 1H, 2H, 2H-perfluorooctyltriethoxysilane (667420, Sigma-Aldrich, USA) in a vacuum dessicator. To fabricate the PDMS devices, PDMS (Sylgard 184, Dow Corning, USA) monomer was mixed with curing agent at 10:1 ratio, degassed, and poured on the mold in a in-house made casing block made of polytetrafluo-

22 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.14: Chemical distribution in the hybrid multiple electric field chip (HMEFC) over 10 hours at 20 µL hr−1. 1 mM 40 kDa fluorescein-dextran with cells was flown in the top channel and 1 mM 40 kDa tetramethylrhodamine-dextran was flown in the bottom channel at 20 µL hr−1. The chemical distribution was stable over a period of 10 hours.

roethylene. After degassing, the PDMS in the casting block was flanked with a piece of 15 mm-thick PMMA block to ensure the exact 4 mm-high final device with surface parallelism. The PDMS was cured under 60◦C for at least 2 hours. After curing, the PDMS devices with the negative impression of the microstructures on the master mold was delaminated and cut into single devices appropriately. Inlet/outlet ports were punched with a 1 mm-diameter biopsy punch (BPP-10F, KAI Group, Japan). The port size around 1 mm was appropriate for biologist-friendly fluid manipulation using a 200 µL pipet tip. Each PDMS device was bonded to a piece of coverglass cleaned to complete the PDMS chip. To ensure surface quality and optical performance, high-precision borosilicate coverglasses (24×60 mm, 170 ±5 µm- thick, No.1.5H, Marienfeld-Superior, Germany) were cleaned in a washing solution with ultrasound (1% TFD4, Franklab, France). The coverglasses were washed thoroughly with ultrapure water, dried, and disinfected with

ultraviolet irradiation prior to bonding to PDMS devices using O2 plasma (AP-300, Nordson MARCH, USA).

PMMA component fabrication 3D microﬂuidic components could be easily and rapidly fabricated using PMMA. In HMEFC, 4 PMMA components were used including the medium inlet reservoir (component A, FIG. S.15.a), outlet/electrical stimulation interface (component B, FIG. S.15.a), top reservoir, and salt bridge connector. The medium inlet reservoir, outlet/electrical stimulation interface, and top reservoir components are composed of four layers of 2 mm PMMA substrates. The designs were created in AutoCAD software and the patterns were cut on a piece of 2 mm casted PMMA

substrate (casted acrylic sheet K, Kanase, Japan) with a CO2 laser cutter (VLS3.50, Universal Laser Systems, USA). 3D microﬂuidic components can be easily fabricated by stacking the laser-cut PMMA pieces and bonding them under high temperature and pressure. The layers were aligned by hand and ﬂanked between two pieces

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

FIG. S.15: The assembled hybrid multiple electric field chip (HMEFC). (a) The exploded view of the components for HMEFC. ds tape: double-sided tape; (b) The 3D rendered model of the the assembled device; (c) The photoimage of the assembled device for experiment. A copper based holder was used to prevent damage to the coverglass. of 2 mm-thick, 100 mm Tempax float glass wafers (Schott AG, Germany) on a force-controlled programmable hot press (G-12RS2000, Orihara Industrial co., ltd, Japan). The PMMA pieces were heated above its glass transition temperature with pressure to form leakage free 3D microfluidic components (118◦C, 500 N, 30 min).

Facile reversible-bonding between PMMA components and PDMS chip A dual energy double-sided tape (85 µm, No. 5302A, Nitto, Japan) was patterned using the CO2 laser scriber. On no.5302A double-sided tapes, a silicone-type adhesive and an acrylic-type adhesive were overlaid on the two sides of a poly(ethylene terephthalate) (PET) substrate. The silicone-type pressure sensitive adhesive adhered to silicone rubber, while the acrylic-type adhesive affixed to plastic, glass, and metal surfaces3,4. The double-sided tape provided an easy and facile way to bond between PDMS and PMMA, an interface typically hard to join. Thus, double-sided tape provides the advantage of high spatial precision of PDMS microfluidics and rapid 3D prototyping of PMMA microfluidics. Using two pieces of patterned double-sided tape, the PMMA component A and B were aligned and affixed to the PDMS chip (FIG. S.15.a). In addition, to avoid breakage and flexing of thin coverglass, a copper holder was used to provide solid support (FIG. S.15.c). The 2 mm-thick copper holder was made of wire electrical discharge machining (wire- EDM) and backed with 0.5 mm PDMS sheet. The PDMS was attached reversibly to the coverglass of the PDMS microfluidic chip.

PMMA top reservoir fabrication To balance the hydrostatic pressure between the inlet and outlet to prevent Poiseuille flow that could displace cells due to hydrostatic pressure, a top reservoir was placed to connect component A and component B. The top reservoir was fabricated by 4 layers of 2 mm-thick PMMA fabricated also with laser cutting and thermal bonding. The top reservoir was made to reversibly bond to PMMA components by affixing 2 mm-thick PDMS paddings using double sided tapes. Openings were cut by a utility knife. Afterward the PDMS part of the top reservoir was affixed to the top of the component A

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.14.948638; this version posted February 20, 2020. The copyright holder for this preprint (whichand Bwas using not certified patterned by peer double-sided review) is the tapes,author/funder, completing who has the granted assembly bioRxiv ofa license the device. to display The the zoomed-inpreprint in perpetuity. view of It theis made components and 3D render of theavailable hybrid under HMEFC aCC-BY-NC-ND were shown 4.0 International in FIG. S.15.b license &. FIG. S.15.c.

References

1C. D. McCaig, B. Song, and A. M. Rajnicek, “Electrical dimensions in cell science”, Journal of Cell Science 122, 4267–4276 (2009). 2H.-F. Tsai, C.-W. Huang, H.-F. Chang, J. J. Chen, C.-H. Lee, and J.-Y. Cheng, “Evaluation of egfr and rtk signaling in the electrotaxis of lung adenocarcinoma cells under direct-current electric field stimulation”, PLoS One 8, e73418 (2013). 3H.-F. Tsai, K. Toda-Peters, and A. Q. Shen, “Glioblastoma adhesion in a quick-fit hybrid microdevice”, Biomedical microdevices 21, 30 (2019). 4C. F. Carlborg, K. B. Gylfason, A. Ka´zmierczak,F. Dortu, M. B. Polo, A. M. Catala, G. M. Kresbach, H. Sohlström,T. Moh, L. Vivien, et al., “A packaged optical slot-waveguide ring resonator sensor array for multiplex label-free assays in labs-on-chips”, Lab on a Chip 10, 281–290 (2010). 5H.-F. Tsai, S.-W. Peng, C.-Y. Wu, H.-F. Chang, and J.-Y. Cheng, “Electrotaxis of oral squamous cell carcinoma cells in a multiple-electric-field chip with uniform flow field.”, Biomicrofluidics 6, 34116 (2012). 6S. Zhao, K. Zhu, Y. Zhang, Z. Zhu, Z. Xu, M. Zhao, and T. Pan, “Electrotaxis-on-a-chip (etc): an integrated quantitative high-throughput screening platform for electrical field-directed cell migration”, Lab on a Chip 14, 4398–4405 (2014). 7K. W. Oh, K. Lee, B. Ahn, and E. P. Furlani, “Design of pressure-driven microfluidic networks using electric circuit analogy”, Lab on a Chip 12, 515–545 (2012). 8D. J. Beebe, G. A. Mensing, and G. M. Walker, “Physics and applications of microfluidics in biology”, Annual Review of Biomedical Engineering 4, 261–286 (2002). 9B. J. Kirby, Micro-and nanoscale fluid mechanics: transport in microfluidic devices (Cambridge university press, 2010). 10T. Kihara, J. Ito, and J. Miyake, “Measurement of biomolecular diffusion in extracellular matrix condensed by fibroblasts using fluorescence correlation spectroscopy.”, PloS one 8, e82382 (2013).